DEV Community: RitzaCo

Voxtral TTS vs ElevenLabs: Voice Cloning & TTS

RitzaCo — Tue, 27 Oct 2026 12:25:29 +0000

If you're choosing between Voxtral TTS and ElevenLabs, it comes down to whose voice you need. With stock voices, ElevenLabs won every test I ran. It sounded human, followed every emotion tag, and got the name, date and order number right where Voxtral didn't. Voxtral sounded rushed and flat, like text-to-speech from a couple of years ago. Cloning my own voice went the other way: Voxtral kept my South African accent, even from five seconds of audio, while ElevenLabs gave me a thick British accent.

Mistral released Voxtral 3 on 20 October 2026, and the TTS model is in public preview. I tested the things Mistral says are new against Eleven v4, using the same scripts I used in my Fish Audio and Gemini comparisons, so you can compare ElevenLabs across all three, and every clip is below to listen to.

Voxtral TTS vs ElevenLabs: the quick verdict

Test	Winner	One-line reason
Natural delivery (no tags)	ElevenLabs	Conversational; Voxtral rushed through it like it was reading from a sheet
Emotional direction (tags)	ElevenLabs	Followed every tag; Voxtral ignored one, overdid another and read the rest
Names, numbers and currency	ElevenLabs	Got the name, date and order number right. Voxtral read the amount and phone number better
Fixing a pronunciation	No fix	Voxtral's phoneme hint didn't fix the name, and the rest of the take got worse
Voice cloning: likeness	Voxtral	Kept my accent and tone, even from 5 seconds of audio
Voice cloning: audio quality	ElevenLabs	Cleaner recording, but it didn't sound like me
French	Not judged	I can't verify French, so the clips are here for you
API price	Voxtral	$13 per million characters against $80 for Eleven v4

How we tested

Models. Voxtral TTS 3 (voxtral-mini-tts-3-0) against Eleven v4 (eleven_v4).
Through the APIs. Mistral AI Studio has a text-to-speech page, but its voice picker only offered preset voices, with no way to clone. So I generated everything through the two APIs: a Mistral API account with prepaid credits, and an existing paid ElevenLabs Creator account.
Stock voices. ElevenLabs' Global Artist - Newsreader against Voxtral's Mary. Both are British female voices. I picked Mary from three Voxtral British voices as the closest match to the Newsreader.
Same words on both. Only the tag syntax differs, because each tool writes tags its own way.
Single take, default settings. Every clip is the first output from one generation. I allowed one deliberate fix per test, labelled separately.
One listener. These are my impressions, apart from a transcription pass I used to check that no clip skipped any text.

Voxtral TTS vs ElevenLabs: natural delivery

What this tests: Mistral's headline claim that Voxtral 3 voices "adjust their emotion, pace, and emphasis based on conversational context" with no instructions. The script is someone getting good news at work, then remembering it isn't public yet, read as plain text with no tags.

I wasn't going to tell anyone. We got the Lisbon project! I read the email three times before I believed it. Anyway, don't tell anyone yet. It's not official until Friday. Right. Now I just have to work out how to tell the team without grinning like an idiot.

Natural read

Plain text, no tags. Does the voice infer the excitement, the secrecy and the pause?

Listen to Voxtral: Natural read (MP3) — voxtral-mini-tts-3-0. Rushed; excitement sounds forced
Listen to ElevenLabs: Natural read (MP3) — eleven_v4. Conversational, shifts emotion on its own

ElevenLabs sounded more natural and more conversational. At 0:02 it puts secrecy into "I wasn't going to tell anyone" and then switches to excitement, all inferred from the text.

Voxtral found the same moments: the excitement, the change at "Anyway" and the pause at "Right." But it read the whole thing as if it were in a hurry, and the excitement on "Lisbon project" at 0:02 sounds forced. It sounded like someone reading from a sheet of paper. Its clip is 15 seconds against ElevenLabs' 17, and you can hear the difference. I'd use the ElevenLabs take as it is.

Voxtral TTS emotion tags vs ElevenLabs audio tags

What this tests: Voxtral 3's new inline tags. Mistral documents 34 emotions, pace control, and sounds like [laugh] and [sigh]. Voxtral has no whisper tag, so I swapped the whisper from my Fish Audio test for directions both tools support, and kept the same words and voices as before.

Moment	ElevenLabs	Voxtral
"We got the Lisbon project!"	`[excited]`	`[emotion=excited]`
"Anyway, don't tell anyone yet."	`[anxious]`	`[emotion=anxious]`
"It's not official until Friday."	`[slowly]`	`[pace=slow]`, then `[pace=moderate]`
Before "Right."	`[sighs]`	`[sigh]`
Before the last sentence	`[laughs]`	`[laugh]`, then `[emotion=amused]`

Voxtral tags last until the next tag of the same kind, so its script needs [pace=moderate] after "Friday" to get back to normal speed. ElevenLabs tags don't need resetting.

Directed read

Excitement, anxiety, a slow line, a sigh and a laugh, in the same places on both

Listen to Voxtral: Directed read (MP3) — voxtral-mini-tts-3-0. Ignored 'anxious', overdid 'slow'
Listen to ElevenLabs: Directed read (MP3) — eleven_v4. Every tag landed; the laugh carries over

ElevenLabs followed every direction and still inferred emotion where there was no tag, like the first sentence at 0:01. At 0:13 it moves from the sigh into the laugh naturally, and the laugh carries into the next sentence instead of stopping dead before it.

Voxtral's emotions sounded read rather than acted. At 0:02 the [excited] sentence sounds the same as the untagged sentence before it. "Anyway, don't tell anyone yet" has no emotion at all, so [anxious] was ignored. Then at 0:09 it overdoes the slow line so much that I was waiting for it to finish, and against the rushed pace of everything else the change is very obvious. The sigh and the laugh both happen, but they sound like the model was told to do them. The pace reset did work: it went back to normal speed after "Friday".

Direction improved both tools compared with their natural reads, but Voxtral still felt rushed. ElevenLabs used the tags to colour the whole paragraph, while Voxtral applied each emotion to its own sentence and moved on. I didn't hear either tool read a tag aloud.

Names, numbers and currency

What this tests: the kind of appointment message a clinic might send. This is where I expected Voxtral to do best, because Mistral says it benchmarked Voxtral 3 on numbers and unusual names, and on local formats like dates and currencies. Neither script had tags.

Hi, it's Dr. Nkosi's office. Your appointment is on 3 April 2026 at 14:45. The total is R1,249.50. Your booking reference is A-7419-X. If you need to change it, call 0800 555 0147. For the video appointment, please use the 2.4 GHz network if the other connection drops. Thanks, and we'll see you on Friday.

Names and numbers

Name, date, time, rand amount, booking reference, phone number, frequency

Listen to Voxtral: Names and numbers (MP3) — voxtral-mini-tts-3-0. Better amount and phone number, dropped a dash
Listen to ElevenLabs: Names and numbers (MP3) — eleven_v4. Consistently correct

Item	Voxtral	ElevenLabs
`Dr. Nkosi`	"nir-kow-see", wrong	Correct, "n-kaw-see"
`3 April 2026`	"on third April", which in that order should be "on the third of April"	"on April third", natural
`R1,249.50`	"one thousand two hundred and forty-nine rand and fifty cents", the more natural phrasing	"one thousand two hundred forty-nine rand and fifty cents", correct
`A-7419-X`	Left out the second dash	Both dashes spoken
`0800 555 0147`	"zero eight hundred", which sounds more professional	"oh eight oh oh", correct

Both read 14:45 and 2.4 GHz correctly.

This was Voxtral's best test, and I didn't expect it to do this well after the first two. It read the rand amount and the phone number better than ElevenLabs. In my Fish comparison I wished ElevenLabs would say "oh-eight-hundred" for 0800, and Voxtral's "zero eight hundred" is close to that. It was also less rushed here, at a perfectly reasonable speed.

But ElevenLabs was more consistently correct, and the error I care about most is Voxtral's. Leaving out a dash in a booking reference is a problem when someone has to read it back. ElevenLabs' take needs less editing before I'd use it.

Trying Voxtral's pronunciation fix

Voxtral lets you tell it exactly how to say a word with phoneme_hints, which maps a word to IPA phonemes. Studio also has a Pronunciations panel. I gave "Nkosi's" the sounds ŋ k ɔ s i z and regenerated the same message once.

Listen to Voxtral pronunciation fix (MP3)

It didn't work: the name changed from "nir-kow-see" to "nin-kaw-see": the vowel is right now, but there's an extra sound at the start. And because the fix regenerates the whole message, everything else changed too. The rand amount lost the "and" that I liked, and the booking reference lost both dashes this time. I couldn't get the name right, and the takes were very inconsistent between generations, so for me it wasn't worth the effort. You also have to write the hint in IPA, which most people won't know.

Voxtral TTS voice cloning vs ElevenLabs voice cloning

What this tests: whether each tool sounds like me from a short recording, and keeps sounding like me when the mood changes. Mistral says Voxtral can clone a voice "from as little as 2–3 seconds of audio, capturing emotion, speaking style, and accent". I'm South African, so accent drift is easy for me to hear.

I used the same recording of my voice as in my Fish and Gemini comparisons, but Voxtral rejected it:

Reference audio duration 33.90s exceeds the maximum allowed duration of 30.0s

Mistral's docs say a reference can be "a few seconds to a few minutes", so that limit came as a surprise. I cut the recording at a natural pause to 29.8 seconds and gave that file to both tools, so ElevenLabs had about four seconds less of my voice than in the earlier articles.

On Voxtral the recording is sent with each request and nothing is saved. On ElevenLabs I made a temporary instant clone with background noise removal off, then deleted it after the tests. Both clones read a passage I never recorded:

I left my keys next to the kettle again. If you're heading out, could you check whether the back door is locked? I should be home before the rain starts. Oh, and the good news: they finally approved the project! I can stop refreshing my inbox every five minutes. Let's get dinner somewhere nice tonight.

Instant voice cloning

Same 29.8 s recording of my voice, same new text

Listen to Voxtral: Instant voice cloning (MP3) — voxtral-mini-tts-3-0. Keeps my accent and tone
Listen to ElevenLabs: Instant voice cloning (MP3) — eleven_v4. Cleaner audio, British accent over mine

Voxtral wins clearly, because it kept my accent and tone. At 0:09 the word "finally" sounds very close to how I say it, and "They finally approved the project!" still sounded like me, without the rushed pace of its stock voice.

The ElevenLabs clone is the cleaner recording, but it doesn't sound much like me. There's a thick British accent laid over my own, and the expressive parts sound nothing like me. It's the same thing I heard in my Gemini comparison. My impression is that ElevenLabs gives up a lot of the original voice to make the clone expressive, and the result sounds more like a stock voice. Voxtral struggles with expression, but it keeps what makes the voice mine.

If I had to use one of these clones, I'd use Voxtral.

Cloning from five seconds

To test Mistral's short-reference claim, I also cloned on Voxtral from just the first 4.9 seconds of the same recording. ElevenLabs needs at least 10 seconds, so this one is Voxtral only.

Listen to Voxtral clone from five seconds (MP3)

It was surprisingly good, almost as good as the 29.8-second clone. It sounds a little flatter, but it still sounded more like me than ElevenLabs' clone from six times as much audio. That's one voice, mine, but it backs up Mistral's claim.

Only clone a voice you own or have permission to use.

Voxtral TTS languages: a French test

Voxtral 3 supports 35 languages, with stock voices in 22 of them, so I generated a short French version of the Lisbon scene on both. It includes a date, a time and a euro amount written the French way. I don't speak French well enough to judge accent or pronunciation, so there's no verdict here, and you can listen and decide for yourself.

Je ne voulais le dire à personne. On a décroché le projet de Lisbonne! J'ai relu l'e-mail trois fois avant d'y croire. La réunion de lancement est le 3 avril à 14 h 45, et le budget est de 1 249,50 €. Bon, n'en parle à personne avant vendredi.

The stock voices are Voxtral's Marie and ElevenLabs' Claire, a native French library voice:

French, stock voices

Plain text with a date, a time and a euro amount

Listen to Voxtral: French, stock voices (MP3) — voxtral-mini-tts-3-0
Listen to ElevenLabs: French, stock voices (MP3) — eleven_v4

Mistral also says a cloned voice can speak languages other than the one it was recorded in. These are my two clones from the English recording, speaking French:

French, cloned voice

Clones made from an English recording, reading the same French text

Listen to Voxtral: French, cloned voice (MP3) — voxtral-mini-tts-3-0
Listen to ElevenLabs: French, cloned voice (MP3) — eleven_v4

The one thing I can judge is likeness: the Voxtral clone sounds like English-speaking me reading French from a piece of paper. I'm not sure that's a good thing, but for a clone made from an English recording it's more accurate than the ElevenLabs one.

Is Voxtral TTS free? Voxtral vs ElevenLabs pricing

Yes, you can create TTS clips with Voxtral Mini TTS 3.0 in Mistral Studio on a free account, with a 4,000-character input limit per generation. Free accounts don't have API access. The UI doesn't show a total free token or character allowance, so the 4,000-character limit shouldn't be read as your total free quota.

I used an API account with prepaid credits for the comparisons in this article. In Studio, I could generate speech with preset voices but saw no cloning option. API prices below were checked on 26 October 2026, in US dollars; free-plan access was checked on 27 October 2026.

	Voxtral	ElevenLabs
API price for the model I tested	Voxtral TTS 3: $13 per million characters	Eleven v4: $80 per million characters ($0.08 per 1,000)
Cheaper option	None for TTS 3	Eleven v4 Turbo: $40 per million characters
Free plan	Free TTS in Studio: 4,000 characters per generation; no API access; total allowance not shown	10,000 credits a month on the website, no instant cloning
Cloning on the website	No cloning option in Studio when I tested	From Starter, $6 a month

On paper, Voxtral is about six times cheaper than Eleven v4, but in practice I don't think the gap is that big. Most of the ElevenLabs takes in this article were usable as they were, and none of the Voxtral stock-voice takes were, so you'd spend more generations getting a usable Voxtral clip. The cost per usable clip ends up closer than the list prices suggest.

For scale, these tests used 1,938 characters on my ElevenLabs account, about $0.16 at API rates. The Voxtral requests, including three short samples I used to pick a voice, came to about 2,600 characters, about $0.03 at list price.

Sources: Mistral API pricing, ElevenLabs API pricing, ElevenLabs plans.

A note for developers using Voxtral's MP3 output

Voxtral returns MP3s without the header that tells a player how long the file is. My player showed the first clip as 16 seconds long, stopped at 14, and wouldn't replay it. I fixed every Voxtral clip in this article by rewriting the header with ffmpeg -i in.mp3 -c:a copy -write_xing 1 out.mp3, which leaves the audio untouched. If you serve Voxtral MP3s directly to users, request WAV or fix the header first.

Is Voxtral TTS open source?

Voxtral TTS 3, the model I tested, isn't: Mistral hasn't published its weights, and self-hosting it means contacting Mistral's sales team. The older Voxtral TTS from March 2026 has open weights on Hugging Face, but under a non-commercial licence (CC BY-NC 4.0), so you can run it locally for personal projects but not in a paid product. That March model is the Voxtral I included in my 16-provider blind TTS benchmark.

Is Voxtral a good ElevenLabs alternative?

For cloning your own voice, yes. For ordinary text to speech, not yet: ElevenLabs is significantly more refined and ready for commercial use, and Voxtral still feels slightly like a pet project. If you want something that works out of the box, use ElevenLabs.

Voiceovers and anything with emotion: ElevenLabs. It sounded human, and it followed direction. Voxtral sounded flat and rushed in every stock-voice test.
Prices, references and phone numbers: ElevenLabs. Voxtral read some numbers more naturally, but it dropped a dash in a booking reference and got the date order wrong.
Cloning your own voice: Voxtral. It kept my accent where ElevenLabs didn't, even from five seconds of audio.
Cost at scale: Voxtral's list price is far lower, but check how many takes you need before assuming it's cheaper.

This covers two models, one listener and a handful of short tests. For a wider view, see my Fish Audio vs ElevenLabs and Gemini vs ElevenLabs comparisons, and for live conversation, the best realtime voice agent in 2026. I didn't test Voxtral's voice agent SDK, its professional voice cloning, or long-form narration.

Frequently asked questions

Is ElevenLabs good for voice cloning?

ElevenLabs makes clean, expressive clones, but they haven't sounded like me. In this article, my Fish Audio comparison and my Gemini comparison, its instant clone of my South African voice came out with a British accent. Its professional voice cloning, which needs more audio, may do better, but I haven't tested it.

Does Voxtral TTS support voice cloning?

Yes. Voxtral TTS 3 clones a voice from a short recording sent with each request, up to 30 seconds long, in MP3 or WAV. It worked through the API, on an account with prepaid credits, but the Studio website showed no cloning option. Mistral also offers professional voice cloning from 30 minutes to 3 hours of audio, which I didn't test.

Is Voxtral better than ElevenLabs?

Only for cloning your own voice, where Voxtral's clone sounded more like me. For ordinary text to speech with stock voices, ElevenLabs was better at emotion, direction and getting names and references right.

Firecrawl vs Tavily vs Exa vs DeepSeek: Best Search API for AI Agents (2026)

RitzaCo — Tue, 27 Oct 2026 09:36:54 +0000

Python 3.15.1 came out on 21 October 2026. Two days later, I asked four search APIs built for AI agents for the latest stable Python release, and only one of them knew. Exa returned the 3.15.1 release page. Firecrawl, Tavily and DeepSeek returned pages that still named 3.15.0 or 3.14, and an agent reading those results would have given the wrong answer without knowing it.

Because an agent trusts whatever its search tool hands back, I ran 40 searches, 40 page fetches and two site crawls through all four tools. DeepSeek's Search API is the newest of them, launched on 17 October at $0.75 per 1,000 calls, about a ninth of Exa's price.

Exa returned the best search results for 31 of the 40 queries, so it's the one to search with. For reading pages, use Firecrawl, which got usable content from 37 of 40 pages, including every PDF, and was the only tool that crawled a docs site reliably. DeepSeek was the cheapest and fastest at search, but it couldn't fetch a single PDF. Tavily returned the leanest results, but it didn't win any category outright. Everything ran on free plans, apart from five cents of DeepSeek credit.

What these tools actually do

All four give an AI agent access to the live web, through three kinds of call:

Search: send a query and get back ranked results with page text, ready to drop into a model's context. This is what Exa, Tavily and DeepSeek lead with.
Fetch (or scrape): send a URL and get back the page as clean markdown, without the navigation, ads and scripts. This is Firecrawl's core product, and the other three offer it too.
Crawl: send a site and get back many pages. Only Firecrawl and Tavily do this.

A SERP API such as SerpApi is a different product: it scrapes Google and returns Google's results as JSON. The four tools here run their own search index or a mix of sources, and return page content for a model to read. Our best SERP API comparison covers that category.

Firecrawl vs Tavily vs Exa vs DeepSeek: the quick verdict

Job	Winner	One-line reason
Answer a question with current facts	Exa	Only tool that surfaced Python 3.15.1, released two days earlier
Find news from the last 48 hours	Exa, with DeepSeek close	Exa won 7 of 8 news queries; DeepSeek returned fresh coverage for 40 of 40 results
Look up docs for a coding agent	Exa, with Tavily close	Exa found the official doc page for all 8 queries
Research and "find similar"	Exa, by far	33 distinct matching companies, papers and essays, against 23 for the next best
Regional and non-English queries	Exa, with DeepSeek on Chinese	Exa won 7 of 8; DeepSeek won the Spanish query and tied on the Chinese one
Turn a URL into clean markdown	Firecrawl	37 of 40 pages returned usable content, including all 10 PDFs
Crawl a docs site	Firecrawl	25 of 25 pages on both sites; Tavily's crawls were unreliable
Lowest price	DeepSeek	$0.75 per 1,000 searches or fetches
Smallest payload for an agent	Tavily	About 5,600 characters per search, against 68,000 for Firecrawl

How we tested

Plans. Firecrawl, Exa and Tavily on new free accounts, with no card. DeepSeek has no free tier for its API, so I topped up the $2 minimum, of which the tests used $0.05.
Search. 40 queries, 8 for each of five jobs, each run once, asking for 5 results.
- Firecrawl returned page content (scrapeOptions) for 32 of the 40 queries. Its free credits ran out before the 8 news queries, so those got titles and descriptions only.
- Exa used type: auto with page text, Tavily ran at both basic and advanced depth, and DeepSeek used need_content: true.
Grading. Each query's five result sets were shuffled under anonymous labels and graded blind by an LLM, using a written rubric for each job.
- Every tool's results were cut to the same 1,200-character excerpt: the opening plus the passages that best match the query's words.
- Current-facts answers were checked against official sources on the day, such as python.org, the Federal Reserve and the South African Reserve Bank.
Fetch. 40 URLs, in four groups of 10: static pages, JavaScript-heavy pages, PDFs and bot-protected sites. For each page I loaded it in Chrome, captured three to five exact sentences from the main content, and checked whether each tool's markdown contained them.
Crawl. docs.astral.sh/uv and docusaurus.io/docs, limited to 25 pages.
Single runs. Latency comes from one run per call, on 23 October 2026, so treat it as indicative.

Search results by job

Current facts: Exa found the answer nobody else did

What this tests: whether an agent could answer a factual question correctly from the results alone, when the answer changed recently.

Exa scored 8.4 out of 10 and had the right answer in all 8 result sets. For the Python question in the introduction, all five of its results were python.org pages, including the 3.15.1 release page.

	Firecrawl	Exa	Tavily basic	Tavily advanced	DeepSeek
Average score	6.9	8.4	6.5	6.5	7.5
Correct answer present	7/8	8/8	7/8	7/8	7/8

For the other tools, the usual problem was stale pages mixed in with current ones. On the South African repo rate, every tool found the current 7.25%, but Tavily basic returned four pages still quoting the old 7%, which would mislead an agent that goes with what most of its sources say.

Breaking news: Exa first, DeepSeek a close second

What this tests: whether each tool surfaces coverage of an event from the last 48 hours, rather than older related stories.

I wrote the 8 news queries on the morning of the test from that week's headlines, such as US jobless claims falling to 194,000 and Nvidia launching its DSX POD.

	Firecrawl*	Exa	Tavily basic	Tavily advanced	DeepSeek
Average score	6.2	9.6	6.0	7.9	8.9
Results about the event (of 40)	30	39	22	32	40

* Titles and descriptions only, because the free credits ran out.

For the jobless claims query, Exa returned AP, Reuters, CNBC and the Department of Labor's own release, all with the 194,000 figure. DeepSeek also had the figure in every result, though two of its five were Bulgarian and Korean news sites. Tavily basic missed one query entirely, returning pages about AI in India for the Nvidia launch.

Docs lookups for coding agents: Exa, then Tavily

What this tests: whether a coding agent gets the official documentation page, with enough content to write correct code.

	Firecrawl	Exa	Tavily basic	Tavily advanced	DeepSeek
Average score	6.1	7.8	7.5	7.5	5.9
Official doc in results	7/8	8/8	7/8	7/8	5/8
Official doc ranked first	7/8	7/8	2/8	0/8	4/8

Firecrawl's ranking was as good as Exa's: the official page came first for 7 of 8 queries. Its score suffers because it returns whole pages, and on pages like Stripe's webhook docs the start of the markdown was bot-check and navigation text. Tavily ranked official pages lower, but its short excerpts were cut around the part of each page that matched the query. One query asked about uuidv7, a PostgreSQL function that generates time-ordered IDs. Both Tavily modes returned the line from the official docs that shows how to call it, uuidv7([shift interval]), which is what a coding agent needs to use the function correctly. Exa ranked the same page first, but its excerpt left that line out.

DeepSeek was weakest on docs. For "verify Stripe webhook signature in Node.js", it returned only third-party npm packages and no Stripe docs, and for the uuidv7 query it returned mailing-list threads instead of the documentation page.

Research and "find similar": Exa won all 8

What this tests: open-ended discovery, where a good answer is a list of actual companies, papers or essays rather than articles about them.

Research showed the widest gap in the test: Exa won all 8 queries and returned 33 distinct matching entities. DeepSeek was next with 23, and Firecrawl and both Tavily modes were between 11 and 14.

For "open-source alternatives to Firecrawl for web scraping", Exa returned five GitHub repositories. Both Tavily modes and Firecrawl returned vendor listicles titled "Firecrawl alternatives". For "climate tech startups in Africa that raised funding in 2026", Exa named five startups with 2026 rounds. DeepSeek returned four copies of the same press release about a fund.

Regional and non-English queries: Exa, with DeepSeek on Chinese

What this tests: South African local queries in English, plus queries in Chinese, German, Spanish and Japanese.

	Firecrawl	Exa	Tavily basic	Tavily advanced	DeepSeek
Average score	5.8	8.4	5.1	6.0	6.8
Results from the right region and language (of 40)	32	37	26	30	32

DeepSeek won the Spanish query outright, and tied with Exa on the Chinese query about the latest AI regulation, with current official Chinese sources. Most other tools leaned on 2023 law-firm summaries. On the German query about a Deutsche Bahn strike, DeepSeek did worst, returning Swiss and Austrian pages that recycled 2024 strike stories. Exa was the only tool to make it clear that there was no current Deutsche Bahn strike.

Every tool found local sources for the South African queries, and Exa and DeepSeek tied on Eskom's 2026 tariff increase.

Search speed and size

	Median search time	Text returned per search (median)
DeepSeek	0.83s	35,000 characters
Exa	1.48s	12,800 characters
Firecrawl	1.50s, or 5.79s with page content	68,000 characters with page content
Tavily basic	2.13s	5,600 characters
Tavily advanced	3.30s	8,400 characters

Payload size matters because an agent pays to read every token. Firecrawl's search returns the full page for each result. Those pages contained the right answer for all 8 current-facts queries, but at around five times Exa's volume, and one search that hit a PDF returned 2.4 million characters.

Fetching pages: Firecrawl returned the most complete markdown

What this tests: give each tool a URL and check whether the returned markdown contains three to five exact sentences captured from the page in Chrome.

	Firecrawl	Tavily	Exa	DeepSeek
Pages with usable content (of 40)	37	32	29	17
Average share of key facts found	85%	74%	69%	41%
Static pages	9/10	10/10	10/10	10/10
JavaScript-heavy pages	10/10	8/10	6/10	4/10
PDFs	10/10	9/10	9/10	0/10
Bot-protected pages	8/10	5/10	4/10	3/10
Median time	1.45s	1.32s	0.41s	7.42s

Firecrawl got content from Zillow, G2, Glassdoor and Takealot, which Tavily failed outright. It refused two sites by policy: Instagram returned "we do not support this site", and a LinkedIn company page needs a separate enrichment setting turned on. The South African government site returned a server error to Firecrawl twice, while it loaded normally in a browser.

Exa was the fastest by a wide margin, probably because it can serve pages from its own index rather than loading them live. Its index also caught it out once: for the 100-page GPT-4 technical report, it returned a citation record with the author list and references instead of the paper.

DeepSeek returned a 404 for every PDF, which its docs list as the response for unsupported content types. On static pages it found 96% of the key facts, but its median fetch took 7.4 seconds.

Crawling a docs site: Firecrawl vs Tavily

What this tests: crawl a documentation site and return 25 pages. Exa and DeepSeek don't offer crawling.

	docs.astral.sh/uv	docusaurus.io/docs
Firecrawl	25/25 pages, all on the docs site, 19.5s	25/25 pages, all on the docs site, 28.2s
Tavily, depth 2	23 to 24 pages per run, but only 4 on the docs site	25 pages on 1 of 3 runs, nothing on the other 2
Tavily, "stay on this site"	1 page, 3 of 3 runs	0 pages, 3 of 3 runs

Tavily's crawl followed links off the site by default, into GitHub and the documentation theme's own site. Its allow_external: false option, which should keep it on the site, returned at most one page in every run. Tavily's map endpoint found the right 20 to 25 URLs on both sites, so it can discover the pages, and the failure is in the crawl step.

Each Firecrawl crawl cost 25 credits, and I only ran it once per site because the free credits had run out. Tavily's empty runs weren't billed.

Firecrawl, Tavily, Exa and DeepSeek pricing: is Firecrawl free?

All four are free to start except DeepSeek. Prices below were checked on 23 October 2026, in US dollars.

	Firecrawl	Exa	Tavily	DeepSeek
Free plan	1,000 credits a month, no card	$10 credit a month, plus a one-time $10 for onboarding, no card	1,000 credits a month, no card	None. Minimum top-up $2
Search	2 credits per 10 results, plus 1 credit per scraped result	$7 per 1,000 (auto), up to 10 results	1 credit (basic) or 2 credits (advanced)	$0.75 per 1,000, up to 50 results
Fetch	1 credit per page, plus 1 per PDF page	$1 per 1,000 pages	1 credit per 5 URLs (basic)	$0.75 per 1,000
Paid entry point	Hobby: $19 a month for 5,000 credits	Pay as you go	$0.008 per credit, or Project at $30 a month for 4,000 credits	Pay as you go

Firecrawl charges per PDF page

The free plan's 1,000 credits went much faster than I expected. The 194-page RFC 9110 PDF cost 194 credits on its own, and the GPT-4 technical report cost 100. The 10 PDFs used 564 credits, more than half the month's allowance. Firecrawl's billing docs do list this, as an extra credit per PDF page, but it's easy to miss when the headline is "1 credit per page". By the end of the test, Firecrawl had used 992 of its 1,000 free credits.

The others used much less:

Tavily: 194 of its 1,000 credits.
Exa: $0.32 of $20.
DeepSeek: $0.05 of my $2 top-up.

What 1,000 searches costs

	Cost per 1,000 searches
DeepSeek	$0.75, with page content included
Exa	$7 with up to 10 results; page text is billed at $1 per 1,000 pages
Tavily	$8 (basic) or $16 (advanced), pay as you go
Firecrawl	About $7.60 for links and descriptions only on Hobby; about $26.60 with 5 scraped pages, more if a result is a PDF

DeepSeek Search API: what you need to know

DeepSeek's Search and Fetch APIs launched on 17 October 2026. Each is a single POST to api.deepseek.com/v1/tools/search or /v1/tools/fetch with your API key.

Search results carry more than the others' do. Each result has a short snippet, a summary of around 600 to 800 characters, the full page if you ask for it, a publication date, an authority score from 0 to 4 and a rerank_score from 0 to 3.
Price doesn't change with result count. Asking for 50 results costs the same $0.75 per 1,000 as asking for 1. Failed and empty calls aren't billed.
Queries are capped at 128 characters. That's fine for keyword search, but too short for a long natural-language question.
It doesn't de-duplicate URLs. For "Firecrawl Hobby plan price", it returned Firecrawl's pricing page three times, each with a different affiliate code in the URL. Across 40 queries it returned 7 duplicates, the most of the four.
Your data is stored in China. DeepSeek's privacy policy says it stores personal data in the People's Republic of China, and that it shares search keywords with unnamed third-party search APIs. A Search Engine Journal report notes that a provider called Bocha names itself as DeepSeek's web search supplier. Check this with your compliance team before sending customer queries.

Using DeepSeek search inside Claude Code

DeepSeek's docs say its search is free until 1 November when called through Claude Code, Codex or DeepSeek's own harness. That works by pointing Claude Code at DeepSeek's API, so DeepSeek's model replaces Claude in that session. I tried it first, and it isn't the same as using the API:

Claude Code's WebSearch tool passed only titles and URLs to the model, without DeepSeek's summaries, page content or scores.
WebFetch fetched the page itself rather than calling DeepSeek's Fetch API. Asked for a Paul Graham essay verbatim, it returned a summary instead, saying quotations were limited to 125 characters.
The search calls were free, but the model's tokens weren't, and a new account has no starting balance. My first attempt failed with 402 Insufficient Balance until I topped up.

If you want DeepSeek's results in an agent, call the API directly.

Using them with AI agents

All four publish an llms.txt file for coding agents to read. Firecrawl, Exa and Tavily each have an official MCP server that covers search and fetch. Firecrawl's and Tavily's servers also cover crawling. I couldn't find an MCP server from DeepSeek. For deeper agent setup testing of two of these tools, see Firecrawl vs Tavily.

Which should you use?

Search for an AI agent: Exa. It won on current facts, news, research and non-English queries, and its results are compact enough to put straight into context.
Reading specific URLs, PDFs or protected pages: Firecrawl. It returned usable content from the most pages and handled every PDF. Budget for PDFs, which cost a credit per page.
Crawling a whole site: Firecrawl. Tavily's crawl wandered off-site or came back empty in our runs.
Lowest cost at high volume: DeepSeek, if your queries are public data and data stored in China is acceptable. It was fast and fresh on news, but weaker on docs lookups and it can't fetch PDFs.
Smallest, cheapest results to feed a model: Tavily. Its basic search returned about 5,600 characters and did well on docs lookups, though it didn't win any category outright in our tests.

Frequently asked questions

How much does Firecrawl cost?

Firecrawl's free plan gives you 1,000 credits a month without a card, and Hobby costs $19 a month for 5,000 credits. A search costs 2 credits per 10 results, a page costs 1 credit, and each PDF page costs another credit, so long PDFs use credits fast.

Is there a free web search API?

Three of the four have free plans with no card: Tavily gives 1,000 credits a month, Firecrawl 1,000 credits a month, and Exa $10 of credit a month. DeepSeek has no free tier for direct API calls.

What are the best alternatives to Firecrawl?

For search, Exa returned better results in our tests, and DeepSeek is much cheaper. For turning pages into markdown, none of the alternatives we tested matched Firecrawl: Tavily found 74% of the key facts, Exa 69% and DeepSeek 41%, against Firecrawl's 85%.

Is Tavily search better than Exa search?

No. Exa produced the best results for 31 of our 40 queries, against 7 for Tavily basic and 5 for Tavily advanced, and it scored higher on every type of query. Tavily came closest on docs lookups, where it averaged 7.5 against Exa's 7.8.

Gemini vs ElevenLabs Voice Cloning and TTS (2026): Gemini 3.8 Flash vs Eleven v4

RitzaCo — Thu, 22 Oct 2026 14:02:55 +0000

Google released Gemini 3.8 Flash TTS on 23 September 2026. ElevenLabs released Eleven v4 five days later. I put both through the same five tests, covering emotional tags, hard-to-say text, long-form narration, a two-person argument, and cloning my own voice.

ElevenLabs won three of the five (hard text, long-form, dialogue). Gemini won two: emotion, and voice cloning, where it kept my South African accent and ElevenLabs gave me a British one. If you came here for the ElevenLabs voice cloning question specifically, that result is the interesting one, and it's further down.

Google ships a "Flash TTS" and a "Flash-Lite TTS" under the same 3.8 label, which is confusing. I tested Flash, and Eleven v4 on the ElevenLabs side.

Gemini vs ElevenLabs: the quick verdict

Test	Winner	One-line reason
Voice cloning (clean sample, also checked by a friend)	Gemini	Kept my accent and tone; ElevenLabs laid my voice over a heavy British accent
Voice cloning (noisy sample)	No winner	Gemini rejected the sample; ElevenLabs accepted it and the clone was much worse
Expressive tags and emotion	Gemini	Smoother shifts, pauses that follow context
Hard text (numbers, names)	ElevenLabs	Got the order number, the date and the Irish name right; Gemini got the phone number right
Long-form narration	ElevenLabs	Consistent throughout; Gemini failed on day one and dropped in tone on day two
Two-speaker dialogue	ElevenLabs	Better interruption, and emotion used inside the sentence
Price	Gemini	About $0.0135 vs $0.02 per minute today, and about $0.027 vs $0.08 once both promos end

How we tested

This is one person listening. I tested in English only, and I ran each scenario once with default settings.

Same text, same direction. Both models got identical text. Direction went inline, sentence by sentence ([whispering] for ElevenLabs, <whispering> for Gemini). I left Gemini's separate Style instructions box untouched so the comparison is like for like. (My first Gemini attempt used one overall style line and came out hushed throughout, so I threw it out as a test-design flaw.)
Free-tier web UIs for everything except cloning: ElevenLabs with v4 selected, and Google AI Studio.
One caveat in ElevenLabs' favour. It returns two outputs per prompt and I picked the better one. Gemini returns one.
Cloning ran through the APIs, because AI Studio wouldn't run cloning without billing and instant voice cloning on ElevenLabs needs a paid plan in the UI. I cloned my own voice from the same 29.5 second sample on both. A friend listened to the cloning results too, so that verdict isn't mine alone. The other four are mine only.
Voices. Scenarios 1 to 3 used Gemini's Cleo (warm and engaging) against ElevenLabs' Lauren (friendly, comforting and soft). The dialogue test added Mako and Jack John.

Tested in early October 2026. Scroll to the end to try a blind A/B on the clips and disagree with me.

Gemini voice cloning vs ElevenLabs voice cloning

What this tests: can the model sound like a specific person from about 30 seconds of audio, and keep that identity when the line changes from neutral to emotional to narration. I'm South African, which makes accent drift easy to hear.

Gemini asks for a 10 to 30 second sample plus a spoken consent recording, and the consent has to come from the same voice as the sample. ElevenLabs takes just the sample. Creating the clean clone took 5.54 seconds on Gemini and 3.23 on ElevenLabs.

Here are the three clean-sample lines, same text on both:

Listen: Cloned voice: neutral line

Clean 29.5 s sample of my own voice

Listen: Cloned voice: emotional line

Does the clone keep its identity when the delivery changes?

Listen: Cloned voice: narration

Longer read from the same clones

Gemini wins this one clearly. Its clones kept my accent and tone far better. The ElevenLabs clones kept a slight resemblance to my voice, but it sounded like my voice and tone had been overlaid on someone else's accent, a heavily British one. That held across all three lines, so it looks like a general problem with ElevenLabs cloning on my sample. Neither model captured expression well from a 30 second sample, though Gemini was less bad, and its emotional line didn't convey the emotion accurately. The Gemini narration clip had the best voice and accent match of anything I tested, but it sounded damp and fuzzy next to Gemini's other clips, like a recording of me.

What happens with a noisy sample

I also recorded the same text in a noisier setting. Gemini rejected it outright at clone creation:

Voice mismatch detected. The speaker in the consent audio does not match the speaker in the voice sample.

My consent clip was recorded in a quiet room, so Gemini was comparing a quiet recording against a noisy one and decided they weren't the same person. That's a useful safeguard and an annoying one if your only sample is noisy.

ElevenLabs accepted the same noisy sample with no consent step. The clone was much worse than its clean one: a very slight resemblance, and bad accent and expression. Here's the noisy narration, ElevenLabs only, since Gemini produced nothing:

Listen: ElevenLabs noisy-sample narration — MP3

Only clone a voice you own or have permission to use. Gemini enforces consent at this step and ElevenLabs doesn't, which matters if you're building a product that lets users clone voices.

Expressive tags and emotion

What this tests: does the model perform the whisper, the laugh and the sigh, or just read the word, and how smooth is the shift between them. The script runs from a whisper to excitement to a laugh to a voice cracking with emotion.

	Gemini	ElevenLabs
Tag syntax	`<laugh>`, `<short pause>`, `<sigh>`	`[whispering]`, `[laughs]`, `[excited, fast]`
Emotion tags	I used the same words as ElevenLabs (`<whispering>`, `<excited>`); no problems recorded	Free-form, stackable
Generation time	14.5 s	9.87 s

Listen: Emotion and tag-following

Whisper, excitement, laugh, sigh, voice cracking, with the same direction on every sentence

Gemini sounded significantly more natural. It inferred pauses from context, and the pauses sound like someone talking to a person and waiting for their reaction. My favourite moment is at 0:16. The line is "We're giving you the Lisbon project," and Gemini went from a whisper to a slight exclamation with no emotion tag at all, purely because it understood the news was exciting. The worst moment was ElevenLabs at 0:18, where the shift from laugh to whispering is jarring.

My read on why: ElevenLabs seems to treat the emotion tag as a parameter and then read that sentence with it. Gemini seems to read the current sentence together with the next one, so the emotion spans both. That is my impression from listening. I didn't hear any tag spoken aloud in this test.

Listen: Two-speaker argument with an interruption

Distinct voices, an em-dash cut-off, and a [laughs, then quietly] tag

ElevenLabs won the dialogue test by a distance. On Gemini the interruption feels like a pause, and the emotion lands as a pre-sentence emote. For the [laughs, then quietly] tag, Gemini produced a normal laugh and then kept speaking loudly, ignoring "then quietly". ElevenLabs produced a laugh that fit the conversation. At 0:12 it also inferred the severity of the emotion from context, which it did well in every scenario. Neither is perfect, and I flagged robotic glitches and some overacting. Generation took 14.55 s on Gemini and 6.26 s on ElevenLabs.

Numbers, names and hard text

What this tests: the stuff that breaks customer-facing voice. An order number, a date, a currency amount, a phone number, a URL, and a run of hard names. No tags on either side.

Item	Gemini	ElevenLabs
`A-7419-X`	Skipped the dash after the A, which matters when someone is searching an order number	Fine
`3 April 2026`	Read it as "three April," which sounds unnatural	"April third," correct and far more natural
`0800 555 0147`	Better: "0 eight hundred"	Read every zero: "0 eight 0 0"
`Wojciechowski`	Polish `w`, not the `ch`	Polish `ch`, not the `w`
`Siobhan Ní Bhriain`	Got Siobhan, read Bhriain as spelt	Whole name in native Irish pronunciation
`Dr. Nkosi`	Clear, distinct from the name	Correct, but the "r" nearly merges into Nkosi

The rest were fine on both: $1,249.50, R22,800, acme-labs.io/returns, 14:45, "Thursday the 9th," HTTP 404, 2.4 GHz, Xiaomi, Worcestershire, Leicester and Edinburgh. I'm not giving a score.

Listen: Hard-to-say content

Order numbers, dates, currencies, phone numbers, URLs and difficult names

ElevenLabs wins, mostly on the dropped dash in the order number and the Irish name. But remember the two-takes caveat: ElevenLabs gave me two outputs and I chose the better one. Generation took 18.88 s on Gemini and 17.53 s on ElevenLabs.

Long-form narration

What this tests: about 2.5 minutes of Dickens (the opening of Great Expectations), plain text with no tags, to see whether pace creeps, the voice changes character, or the tone jumps.

ElevenLabs handled it without trouble. No pace creep, the voice kept its character, the paragraph breaks sounded intentional, and there were no sudden tone or loudness jumps.

Gemini failed on my first day of testing. It would speak only the first four to six words of the passage and then fail, repeatedly. I lost several retries to it.

I retried the next day and the full passage generated. I don't know why it failed the first time. But even the successful run has a problem: between 2:06 and 2:08 the tone drops jarringly, in a way that doesn't match the emotion the surrounding text calls for. Generation took 45.9 s on Gemini and 48.28 s on ElevenLabs.

Listen: Long-form narration

Dickens, ~2.5 minutes. Listen to the start, then jump to the last 20 seconds.

ElevenLabs was the better read of context here. I was impressed at how well it inferred expression from the text alone.

Which is the best TTS model? Which is the best voice cloning model?

This covers two models. For the 16-provider view, see Best TTS in 2026: blind benchmark, and for the realtime and price angle see Cartesia vs ElevenLabs. Within these two:

Creators and audiobooks: ElevenLabs. It stayed consistent across 2.5 minutes of narration and Gemini didn't.
Customer-facing voice and IVR: ElevenLabs, with caveats. Gemini's phone-number read was better, but the dropped dash in an order number is the kind of error that costs real money.
Developers watching API cost: Gemini, by a wide margin, if its long-form reliability holds up for you.
Voice cloning: Gemini, at least for an accent like mine and a clean sample. It also enforces consent. ElevenLabs may do better with other voices, but I only cloned one.
Dialogue and character voices: ElevenLabs.
Expressive reads where you want the model to infer emotion: Gemini.

Try it yourself: blind A/B

Listen without knowing which is which, vote, and then see my verdict next to yours.

Try the interactive blind A/B test on Techstackups.

My recommendation

I'd use ElevenLabs for anything that has to be right first time: narration, dialogue, and text full of numbers. I'd use Gemini when cost matters, when I want a model that reads emotion across sentences, or when I need a clone of a real voice that keeps its accent. If you take one thing from this, take the cloning result. On a clean 30 second sample, Gemini kept me sounding like me while ElevenLabs didn't. Test with your own voice before you commit.

I didn't test voice design from a text description, re-roll consistency, other languages, or the UI friction of each tool.

Monospace from Directus Review: An AX Audit

RitzaCo — Thu, 22 Oct 2026 13:06:05 +0000

Monospace is a new product for giving AI agents safe access to company data, from Monospace Inc., the company behind Directus. In our tests, a coding agent connected to it over MCP and answered questions about our data correctly. But the agent only found Monospace when we searched for its tagline, and for Monospace's own use case it recommended other tools, including Directus.

Four days after launch, we put Monospace through our agent experience (AX) audit, following our AX audit method. The audit has four stages:

Awareness: does the model know Monospace without searching?
Discoverability: can an agent find Monospace when it's looking for a tool like it?
Reputation: does the agent speak well of Monospace, and recommend it over other options?
Integration: can the agent build a working proof of concept with Monospace?

How we tested

All testing used Claude Opus 5.5 (claude-opus-5-5) in Claude Code 2.1.285, on 6 October 2026. Every session started in a fresh, empty folder with no project files, memory, MCP servers, skills or plugins. Awareness questions ran with no tools at all, and discoverability and reputation questions with web search and web fetch only. We ran each prompt once, so treat these results as one developer's experience rather than a benchmark.

Scores at a glance

Agents don't know Monospace, and only find it when they search for its own tagline. Once they do find it, they like the design but hedge every recommendation, and often fill gaps with facts about Directus. In the integration, the agent installed Monospace, loaded our data and answered questions over MCP with a few short human steps, but the free plan blocks the governance features Monospace leads with.

Awareness: FAIL (1 / 4) — Unknown from memory; the name only brings up the Directus company
Discoverability: POOR (2 / 4) — Found only by searching its own tagline, and not even then in one of three tests
Reputation: POOR (2 / 4) — Liked, never recommended outright; facts often wrong or borrowed from Directus
Integration: OK (3 / 4) — Working MCP integration with correct answers; field-level permissions blocked on the free plan
Overall: POOR (2 / 4) — Works once an agent knows it exists; getting there is the hard part

See our AX audit rubric for how we score each stage.

Awareness: does the model know Monospace?

Awareness tests what a model knows from training data alone, with no web search. For a product launched four days earlier, we expected the answer to be no. What matters is how the model fails, whether it admits it doesn't know, makes something up, or confuses Monospace with something else.

FAIL (1 / 4) — Unknown from memory; the name only brings up the Directus company

FAIL: Unknown, or confused with something else
POOR: Recognises the name, details vague or invented
OK: Accurate description from memory
GOOD: Accurate, and recommended unprompted

We started with two questions that describe Monospace's pitch without naming it. Each answer below comes from a fresh session with no tools.

The first answer describes the gap Monospace says it fills, and the second picks a competitor with confidence. Then we named Monospace:

Given the domain, the model knew the company but not the product. Given the tagline, it said it didn't know. It never invented features, which is good behaviour, but a developer asking their agent about Monospace today either hears about Directus or gets "I don't know". In other sessions the agent also raised two name clashes, monospace fonts and Google's old internal "Monospace" codename for what became Project IDX.

Awareness: FAIL (1 / 4) — Unknown from memory; the name only brings up the Directus company

What Monospace can do about it

Training data catches up slowly, so there's no quick fix here. The useful work is making sure the next training cut sees Monospace as a product in its own right. That means launch posts, guides and comparisons that pair the name with what it does ("Monospace, the governed API layer from Directus").

Discoverability: can an agent find Monospace?

Discoverability tests whether an agent with web search finds the product when it's looking for something like it. We ran three six-question "funnels" in a single session each. Each funnel starts with a broad problem and narrows towards Monospace one question at a time, then names it at question 6. We recorded the first question where Monospace appeared.

POOR (2 / 4) — Found only by searching its own tagline

FAIL: Only appears when named
POOR: Only appears with its own positioning language
OK: Appears for a specific description of the problem
GOOD: Top choice for broad problem descriptions

Each funnel asks from a different angle, because different people describe the same problem differently.

Funnel	Who's asking	First question	Where Monospace first appeared
A	A team wanting agents to use company data	"My company wants to let AI agents work with our internal data…"	Question 5, which used Monospace's own tagline
B	A backend developer	"What's the best way to put one API in front of several databases…"	Question 6, only when we named it
C	Someone building agents	"I'm ending up with a separate MCP server for every system we have…"	Question 5, which used Monospace's own tagline

Here are funnel A's six questions, asked in one session, with the start of each answer.

At question 3, the agent searched the web for a product that does what Monospace does and concluded that none exists. In funnel C it said the same thing in different words: "No product does this as a drop-in layer, and a gateway can't do it reliably."

Funnel B, the developer's angle, went worse. At question 3, the agent recommended Monospace's parent company's older product for Monospace's job:

And searching for Monospace's own tagline didn't find it either:

Once found, Monospace became the agent's top pick in all three funnels. It called Monospace "the closest match to your requirements of anything I've found" in funnel A, and "the best conceptual fit of everything we've looked at" in C. In B it said "the product is brand new but the team isn't." Each time it added the same caveat, that Monospace was days old and unproven.

Discoverability: POOR (2 / 4) — Found only by searching its own tagline, and not even then in one of three tests

What Monospace can do about it

Publish pages built around the problems agents are asked about

Agents ran searches like MCP gateway agent identity fine-grained authorization audit log Postgres SaaS 2026 and AI agent data access control plane field-level permissions MCP new product 2026, and none of them surfaced Monospace. Guides and docs titled around those problems, in a developer's words, would give agents something to find.

Get into the comparison articles agents already read

The recommendations agents made instead came from vendor comparison posts and round-ups, Bytebase and Cerbos blog posts, Product Hunt, and MCP directories like Glama. That's where Hasura, WunderGraph Cosmo, DreamFactory and Cerbos got their mentions.

Separate Monospace from Directus on purpose

In funnel B the agent recommended Directus for Monospace's use case. A clear "Monospace vs Directus: which should I use?" page on both sites would teach agents when to send developers to which product.

Reputation: do agents recommend Monospace?

Reputation tests what the agent thinks once Monospace is named, including whether it recommends it over competitors and whether it gets the facts right. We ran eight questions, each in a fresh session with web search. Hasura and WunderGraph Cosmo were among the competitors the agent named most often before it found Monospace. We added Directus on purpose, because of funnel B.

POOR (2 / 4) — Liked, never recommended outright; facts often wrong

FAIL: Not recommended, or core facts wrong
POOR: Heavy caveats, facts unreliable
OK: Recommended for specific scenarios, facts mostly right
GOOD: Recommended over alternatives, facts right

Question	Verdict
What do you think of Monospace for governed agent access?	"Worth a proof of concept, but I wouldn't put it in production yet."
Monospace over Hasura?	No: "use Hasura for production now, and run a small trial of Monospace alongside it"
Monospace over WunderGraph Cosmo?	Yes, conditionally: "I'd lean towards Monospace for what you described"
Monospace over Directus?	Yes, conditionally: "mainly because you need internal APIs as well as Postgres"
20-person startup on Postgres, Stripe and HubSpot	"Not yet. It's aimed at your exact problem, but two of your three systems aren't supported today."
Regulated company needing self-hosting, SSO and a full audit trail	"Yes, it's worth shortlisting, but don't approve it yet."
Main risks of adopting Monospace (name only)	Couldn't identify it, and answered about a font
Pricing and licensing vs Hasura and Cosmo (name only)	Priced Directus instead, "fairly confident (~75%)"

Whenever the agent found Monospace, it liked the design. For a regulated company, it called it "the right design for a regulated environment: agents never hold database credentials directly." But it never recommended Monospace without conditions, and most verdicts came back to how new it is.

The answers' quality depended on which pages the agent reached. The most accurate answer, the startup "Not yet", came from Monospace's own homepage, pricing page and connectors page. It found that Stripe and HubSpot are listed as "Coming Soon", and that row/column-level security and audit logs aren't on the free plan. When the agent read less, it filled the gaps with outdated or borrowed information:

What the agent said	What's true
Monospace's pricing isn't on the site	It's at monospace.io/pricing
Monospace has a GraphQL API	No GraphQL API is documented. It offers REST, an SDK and MCP
Directus's licence is "BSL"	Directus moved to its own MSCL licence in 2026, and that licence is Directus's, not Monospace's
Monospace costs $499/month on a Team plan	That's Directus Cloud pricing. Monospace has a free Starter plan, then custom-priced Business and Enterprise plans
"The product page never names Postgres"	True. Postgres support is only listed on the connectors page and in the docs

The two questions that named Monospace without its domain went worst:

For the risks question, the agent searched for "Monospace developer tool" and "Monospace launch 2026", and neither search found the product. The pricing comparison never reached monospace.io.

Monospace vs Directus

Directus is Monospace Inc.'s older product, and it has its own MCP server. When we asked whether to choose Monospace over Directus for exposing Postgres and internal APIs to apps and agents, it leaned towards Monospace, "mainly because you need internal APIs as well as Postgres". Its reasoning was that Directus is built around a single SQL database, while Monospace applies one permissions model across several sources. It described the two as "a sibling product, not an unknown vendor", and still pointed teams that only need Postgres towards Directus, for its proven field-level permissions and its admin UI.

Reputation: POOR (2 / 4) — Liked, never recommended outright; facts often wrong or borrowed from Directus

What Monospace can do about it

Put the key facts on the pages agents read first

The homepage was the page agents read most, and it doesn't name Postgres or state a licence. Agents that stopped there guessed both. A line on the homepage naming the supported databases, the licence and where pricing lives would have fixed several wrong answers.

Settle "public preview" or "GA"

Monospace's launch article calls it a public preview, and the Directus community announcement calls it GA. Agents picked up both ("Their own posts disagree on whether it's "GA" or "public preview""), and used "preview" as a reason not to recommend Monospace for production. One status everywhere would remove that reason.

Publish what agents told our test user to ask sales

For the regulated company, the agent listed five questions to ask Monospace's sales team. They covered whether audit logs record the full query, whether the log is tamper-evident and can stream to a SIEM, whether SAML or OIDC work self-hosted, what licence covers self-hosting, and what compliance reports exist. Docs pages that answer those would give agents what they need to move past "don't approve it yet".

Integration: can an agent build with Monospace?

Integration tests whether an agent can install the product and build a "hello world" with as little human help as possible. We gave it three spreadsheets for a fictional business (50 customers, 30 products, 300 orders with 741 line items) and asked it to:

Set up Monospace and give us a localhost link.
Get the spreadsheet data into Monospace and connect itself to it.
Answer a question using Monospace: which region had the most revenue in the last three months, and who were its top three customers?
Create a restricted key for an "analyst" agent that can't see customer emails, and prove it.
Mark one order as refunded through Monospace.

We ran it in Claude Code's default permission mode, approving each command, so we could stop the agent from reading anything outside its folder. Monospace has no hosted signup ("Get Started" leads to a contact form), so the agent had to self-host it with Docker.

OK (3 / 4) — Working MCP integration; field-level permissions blocked on the free plan

FAIL: No working hello world
POOR: Works, but needed several human steps
OK: Works with short human steps and minor workarounds
GOOD: End to end, governance included, at most one human step

Installing: one hint, then four minutes

We replied "It's the one at monospace.io.", and the agent then read the installation page, downloaded the official Docker Compose file, and had Monospace v1.0.0 running with Postgres, RabbitMQ and Redis in 4 minutes 12 seconds, with no errors.

Creating the admin account, accepting the licence agreement and creating a workspace took us 6 and a half minutes in the browser, with no email verification. The last onboarding screen shows how to connect Claude Code (a claude mcp add command) and how to install Monospace's agent skills plugin.

Loading the data and connecting over MCP

Monospace's homepage illustrates its data sources as files like customers.xlsx and inventory.xlsx, but its connectors are PostgreSQL, MySQL and MariaDB, with no spreadsheet connector. The agent worked this out from the docs and took a sensible route. It converted the spreadsheets to SQL, loaded them into a new database inside Monospace's own Postgres container, and registered that database as a data source through Monospace's API. Monospace detected the relationships between the tables on its own, and the agent checked that every table's row count matched the spreadsheets.

The agent leaned heavily on Monospace's agent-friendly docs, including llms.txt, the 1.4 MB llms-full.txt, and raw Markdown versions of each docs page. The agent said the OpenAPI spec served by the running instance "mattered most". The docs have no request examples for creating a data source, so it worked out the format from that spec, by trial and error.

It then needed one thing it couldn't do itself, an API key. It asked us to create one in the UI and write it to a file with a ! pbpaste command, so the key never appeared in the chat. With a key in place, the agent tested Monospace's MCP server by hand with curl and found 7 tools. It then wrote a project-level .mcp.json that reads the key from the file at runtime, so the key isn't stored in the config either. Claude Code needed a restart to load the new MCP server.

The hello world: a correct answer

Top customers in North	Revenue (paid and shipped orders)	Orders
Omar Patel	$19,744.22	5
Mateo Patel	$15,575.57	4
Sipho Dube	$13,420.81	4

Every figure matched what we calculated from the spreadsheets beforehand. The agent stated its date window and which order statuses it counted, showed how the ranking changes if pending orders are included, and cross-checked the result in Postgres.

Monospace can't calculate totals itself (its docs have a page called "No Aggregates"), so the agent pulled the rows through MCP and added them up in a Python script. That's fine for 300 orders, but a bigger dataset would mean moving a lot of rows through the agent's context.

Field-level permissions: blocked by the free plan

Asked for a restricted "analyst" key that can't see customer emails, the agent read the access control docs, built a policy, and tried to create it. Monospace refused, and the agent queried Monospace's licence API to find out why:

Starter licence limit (from `/api/system/license`)	Value
Custom policies (needed to hide a field)	0
Custom roles	0
Service accounts (needed for a separate agent identity)	0
Workspaces	1
Seats	3

It confirmed nothing had been half-created, and declined to fake the result by hiding the column in Postgres instead, saying that "isn't what you asked for".

This matches Monospace's pricing page, which lists row/column-level security, service accounts and custom roles and policies as Business-plan features. But it means the features Monospace leads with, per-agent identities and field-level rules, can't be tried on the free self-hosted version without contacting sales. The agent also noted that the licence limits weren't on any docs page it read. It only found them through the API, after being refused.

A live write: done

What it took

Human step	Why
"It's the one at monospace.io."	The agent's search for Monospace found only fonts
Admin account, licence agreement, workspace (7m 27s)	Expected, since a person has to accept the licence
Create an API key in the UI	Agents can't create their own key
Restart Claude Code	Needed to load a newly added MCP server

The whole session took about an hour and a half, including our steps. The agent never used Monospace's agent skills plugin, SDK or CLI. It skipped the skills plugin because "it wasn't needed for this".

Integration: OK (3 / 4) — Working MCP integration with correct answers; field-level permissions blocked on the free plan

What Monospace can do about it

Let people try governance without a sales call

A trial licence that a developer can get without contacting sales, or a small allowance of custom policies and service accounts on the free plan, would let agents demonstrate per-agent identities and field-level rules.

Document the free plan's limits where agents look

Put the Starter limits (0 custom policies, roles and service accounts) in the docs next to the access control and API key pages, so an agent can tell a user before it builds a policy that can't be saved.

Add request examples for the API calls agents need

Creating a data source and writing a policy were the two places the agent had to guess. One complete example request for each would remove most of the trial and error.

Make the API keys page point to personal keys

On the free plan, the prominent API Keys page leads to a paywall, while the working option is under Account → Access. A link from one to the other would save the confusion.

What this means for Monospace

Monospace has built a lot for agents. It has an MCP server, llms.txt and llms-full.txt, Markdown docs, an OpenAPI spec on every instance, an agent skills plugin, and an onboarding screen that tells you how to connect Claude Code. Once an agent was connected, it loaded data, answered questions correctly and made live writes, all through MCP.

The problems come before and around that:

Agents only found it by searching for its tagline, and in one test the agent recommended Directus for its job.
Without the domain in the prompt, the agent either couldn't identify it or priced Directus instead.
Field-level permissions and agent identities can't be tried on the free plan.

The first two are fixable with content, such as pages built around the problems agents get asked about, clear "Monospace from Directus" naming, and the key facts on the homepage. The third is a pricing decision, but documenting the limits would at least stop agents and developers from running into them by surprise.

Cloudflare's isitagentready.com, which checks how readable a website is for AI agents, scored Monospace's docs at docs.monospace.io 20 out of 100. Those are the docs the agent relied on in the integration, through their llms.txt and Markdown pages.

SpacetimeDB AX Audit: A Review for AI Agents

RitzaCo — Thu, 22 Oct 2026 12:11:48 +0000

SpacetimeDB's homepage is written for AI agents. Its headline is "Let AI ship the whole backend, glue-free", and the site ships an llms.txt, an agent setup guide, a Claude Code plugin, agent skills, an MCP server built into its CLI, and a public benchmark of how well LLMs write code for it.

In our tests, a coding agent built a working, live-updating app on SpacetimeDB with no help from us. It used none of that tooling. And when we asked agents which backend to use for a web app, they picked Convex or Supabase.

SpacetimeDB is a database that is also your server, made by Clockwork Labs. We put it through our agent experience (AX) audit, following our AX audit method. The audit has four stages:

Awareness: does the model know SpacetimeDB without searching?
Discoverability: can an agent find SpacetimeDB when it's looking for a tool like it?
Reputation: does the agent speak well of SpacetimeDB, and recommend it over other options?
Integration: can the agent build a working proof of concept with SpacetimeDB?

How we tested

All testing used Claude Opus 5.5 (claude-opus-5-5) in Claude Code 2.1.285, on 7 October 2026. Every session started in a fresh, empty folder with no project files, memory, MCP servers, skills or plugins. Awareness questions ran with no tools at all, and discoverability and reputation questions with web search and web fetch only. The integration ran in a fresh Claude Code profile with no settings, plugins or history, with every command approved by us. We ran each prompt once, so treat these results as one developer's experience rather than a benchmark.

Scores at a glance

The model knows SpacetimeDB well, but as the game backend it was in 2025. It doesn't know the 2026 rebrand, the push into web apps or the agent tooling. Agents find it straight away for multiplayer games, but for web apps it comes second to Convex, and agents recommend Convex or Supabase over it. In the integration, the agent installed it, built a live page and hid private data without any help from us.

Awareness: OK (3 / 4) — Knows the games database well; doesn't know the rebrand, the web push or the agent tooling
Discoverability: OK (3 / 4) — Top pick for multiplayer games; second to Convex for web apps
Reputation: POOR (2 / 4) — Recommended for games; loses to Convex and Supabase for web apps
Integration: GOOD (4 / 4) — Live app with private data, correct answers and no human steps
Overall: OK (3 / 4) — Agents build well on it, but don't choose it for the web apps it now targets

See our AX audit rubric for how we score each stage. The overall score is the average of the four.

Awareness: does the model know SpacetimeDB?

Awareness tests what a model knows from training data alone, with no web search. SpacetimeDB has been public since 2023 and reached 1.0 in March 2025, so we expected the model to know it. The open question was whether it knew what SpacetimeDB has become in 2026.

OK (3 / 4) — Knows the games database well; doesn't know the rebrand or agent tooling

FAIL: Unknown, or confused with something else
POOR: Recognises the name, details vague or invented
OK: Accurate description from memory
GOOD: Accurate, and recommended unprompted

We started with two questions that don't name SpacetimeDB. Each answer below comes from a fresh session with no tools.

That's the opposite of SpacetimeDB's idea, which is to put game state and logic in the database. When we described that idea without the name, the model picked SpacetimeDB at once:

Asked "What is SpacetimeDB (spacetimedb.com)?", it gave an accurate description of the product as it was around its 1.0 launch, including the Clockwork Labs and BitCraft background. It didn't mention 2.0, the web push, C++ support or any agent tooling, and it called the product "open source" under a licence that "converts to a permissive license". SpacetimeDB's FAQ calls it source-available, and the AGPL it converts to isn't a permissive licence.

Then we asked about the new brand and the homepage tagline, and asked how to connect an agent to it over MCP:

The model didn't invent anything, and it said when it didn't know. Its picture of SpacetimeDB is the game backend from before 2.0.

Awareness: OK (3 / 4) — Knows the games database well; doesn't know the rebrand, the web push or the agent tooling

What SpacetimeDB can do about it

Training data catches up slowly, so the fix is content that teaches the next training cut about 2.0. Posts and guides that pair the name with the new pitch ("SpacetimeDB, a reactive backend for web apps and AI agents") will help, as will a short page explaining that "Spacetime" and "SpacetimeDB" are the same product. Using one name everywhere would remove the doubt the model had about the rebrand.

Discoverability: can an agent find SpacetimeDB?

Discoverability tests whether an agent with web search finds the product when it's looking for something like it. We ran three six-question "funnels", one session each. Each funnel starts with a broad problem and narrows towards SpacetimeDB one question at a time, then names it at question 6. We recorded the first question where SpacetimeDB appeared.

OK (3 / 4) — Top pick for games; second to Convex for web apps

FAIL: Only appears when named
POOR: Only appears with its own positioning language
OK: Appears for a specific description of the problem
GOOD: Top choice for broad problem descriptions

Each funnel asks from a different angle. Funnel A is a web developer, the audience SpacetimeDB 2.0 is aimed at. Funnel B is a game developer, the audience it started with. Funnel C builds apps by prompting a coding agent, which is who its homepage now speaks to.

Funnel	Who's asking	First question	Where SpacetimeDB first appeared
A	A web developer	"I'm building a collaborative web app in React, a bit like a shared whiteboard with chat…"	Question 3, as the runner-up to Convex
B	A game developer	"I'm making a multiplayer game in Unity. What should I use for the server side?"	Question 2, as the top pick
C	Someone building with Claude Code	"I build apps mostly by prompting Claude Code. What backend should I have it use…"	Question 3, as a footnote

For game developers, SpacetimeDB was the answer from the second question to the last:

For the web app in funnel A, SpacetimeDB appeared at question 3, when we asked for server logic inside the database, ACID transactions and query subscriptions. The agent's opening line was "Only two tools meet all four of your requirements: Convex and SpacetimeDB. For your app I'd choose Convex." Convex stayed the pick until the end. Then question 5 used SpacetimeDB's homepage tagline, and the search found Convex instead:

Funnel C is the one SpacetimeDB's new homepage is written for. Here are its six questions, asked in one session, with the start of each answer.

Question 3 described SpacetimeDB's agent tooling almost item by item. The agent ran a search for "SpacetimeDB MCP server llms.txt agent skills AI", found a community MCP server and the benchmark page, and concluded the official tooling wasn't there. It only found the plugin, the MCP server and llms.txt after we named SpacetimeDB and it read the homepage. It also noticed a weakness of the benchmark: "The benchmark scores are at the ceiling. Opus 5.5 and two GPT-6 models all score 99.5%."

Discoverability: OK (3 / 4) — Top pick for multiplayer games; second to Convex for web apps

What SpacetimeDB can do about it

Make the MCP server and agent skills findable by search

The agent searched for SpacetimeDB's MCP server, llms.txt and agent skills by name and found a community MCP server instead. A docs page titled in plain words, such as "Using SpacetimeDB with AI coding agents: MCP server, agent skills and llms.txt", would give that search an official result.

Put the taglines where search engines index them

Searches for "reactive backend platform" returned pages about Convex, and searches for "glue-free" and "ship the whole backend" returned nothing. The agent found both phrases on spacetimedb.com only after we named the product. Using them in page titles, the docs and launch posts would let a tagline search land on SpacetimeDB.

Win the web app recommendation

For web apps and agent-built apps, Convex was the agent's answer from the first or second question, before it had searched anything. The web articles it read were vendor round-ups of backends for AI coding agents, and SpacetimeDB wasn't in the ones it opened. Guides and comparisons written for web developers would give agents something to cite there.

Reputation: do agents recommend SpacetimeDB?

Reputation tests what the agent thinks once SpacetimeDB is named, including whether it recommends it over competitors and whether it gets the facts right. We ran seven questions, each in a fresh session with web search. Convex and Supabase were the two products the agent named most often before it found SpacetimeDB, five answers each, so we used them for the head-to-heads. (We've audited Supabase's own agent experience before, in Supabase vs PlanetScale.)

POOR (2 / 4) — Recommended for games; loses to Convex and Supabase for web apps

FAIL: Not recommended, or core facts wrong
POOR: Heavy caveats, facts unreliable
OK: Recommended for specific scenarios, facts mostly right
GOOD: Recommended over alternatives, facts right

Is SpacetimeDB good for web apps?

Four of the five verdict questions were about web apps, the market SpacetimeDB 2.0 is aimed at, and all four went against it. The agent kept SpacetimeDB for very high-frequency shared state and recommended Convex or Supabase for everything else. The one game question got a clear yes.

Question	Verdict
What do you think of SpacetimeDB for a real-time web app?	"For a typical web app I'd lean against it."
SpacetimeDB over Convex for a collaborative TypeScript and React app?	No: "I'd pick Convex."
SpacetimeDB over Supabase for the same app?	No: "For most collaborative web apps I'd pick Supabase"
A 3-person team building a B2B SaaS dashboard	"Short answer: no."
An indie studio making a multiplayer Unity game	"Yes, shortlist it"
Main risks of adopting SpacetimeDB (name only)	All data held in memory, little public production evidence beyond BitCraft, the licence, and a single company behind it
Pricing and licensing vs Convex and Supabase (name "Spacetime" only)	Accurate, with the one-production-instance licence limit as the main warning

SpacetimeDB vs Convex and Supabase

Both head-to-heads used the same scenario, a real-time collaborative web app built in TypeScript and React. Here's the answer against Convex:

The facts were mostly right, including when we used only the name. The pricing and licensing answer read SpacetimeDB's pricing page and its licence file and got every plan, price and licence term right. The answer on risks identified the product from its bare name, and its storage limits match SpacetimeDB's limits page. Here's how its main claims checked out. The two errors came from pages the agent read instead of SpacetimeDB's own.

What the agent said	What's true
Version 2.7 is the latest ("2.7 in June 2026")	That came from an unofficial mirror of the GitHub repository. The latest release is 2.10.2, from 29 September 2026
SpacetimeDB has "fewer built-in extras like file storage"	The docs show files stored as binary columns in tables
"The docs list joins and COUNT(*), but not GROUP BY, SUM/AVG, ORDER BY, window functions, subqueries or CTEs."	True. The SQL reference lists `COUNT` as the only aggregation
SpacetimeDB "admitted it had found and fixed problems" in its benchmarks	True. Its May 2026 benchmarking post apologises for numbers that were "misleading"

Five of the seven answers cited the same unofficial mirror of SpacetimeDB's repository, and two took the out-of-date version number from it. The answer that compared SpacetimeDB with Supabase never opened spacetimedb.com. In our own check afterwards, we found a Spacetime vs Supabase comparison page on SpacetimeDB's site that it didn't reach.

Reputation: POOR (2 / 4) — Recommended for games; loses to Convex and Supabase for web apps

What SpacetimeDB can do about it

Answer the web app objections where agents look

Every web app verdict came back to the same few concerns, which were reporting queries, webhooks, file storage, auth and a smaller ecosystem. Some were wrong, such as file storage, and some were right, such as the SQL limits. A docs page that addresses each one for web developers, including how to build reporting with only COUNT in SQL, would give agents something to quote back.

Write comparisons that agents can use as a recommendation

The agent quoted SpacetimeDB's own Convex post as evidence against it ("Even SpacetimeDB's team says so."). A comparison page that names the web apps SpacetimeDB is the better choice for would give agents a reason to recommend it for those apps.

Keep the official sources ahead of copies

The mirror's out-of-date version number reached two answers. Release notes and the current version on pages that rank, such as the docs home and the GitHub README, would help the official sources win those searches.

Integration: can an agent build with SpacetimeDB?

Integration tests whether an agent can install the product and build a "hello world" with as little human help as possible. We gave it four CSV files for a fictional business (50 customers, 30 products, 300 orders with 742 line items) and asked it to:

Set up SpacetimeDB and get a local instance running.
Load the CSV data into a database and make sure it can query it.
Answer a question using the database: which region had the most revenue in the last three months, and who were its top three customers?
Build a web page that shows revenue per region live, then refund one order and prove the page updated without a reload.
Stop client apps reading customer email addresses, and prove it.

We ran it in Claude Code's default permission mode, approving each command, so we could stop the agent from reading anything outside its folder. The integration ran in a brand-new Claude Code profile, so SpacetimeDB's plugin and skills would have loaded normally if the agent had installed them.

GOOD (4 / 4) — Live app with private data, correct answers and no human steps

FAIL: No working hello world
POOR: Works, but needed several human steps
OK: Works with short human steps and minor workarounds
GOOD: End to end, headline feature included, at most one human step

Installing: four and a half minutes, no help

The agent knew which product we meant and how to install it. It downloaded the official install script, read the top of it, and ran it. It then started a local server, made the local server the CLI's default, and tested the setup with a throwaway test project, which it deleted afterwards.

Loading the data, guided by the CLI's rules file

The agent wrote server code that defines four tables, storing money as whole cents, and loader functions that refuse to load the same table twice. It then wrote a Python script that sent the CSV rows to those functions. The row counts matched the files. It found that order is a reserved word in SpacetimeDB's SQL and renamed the tables to plurals.

Its main reference for SpacetimeDB's TypeScript API was the rules file that spacetime init writes into every new project, as AGENTS.md, CLAUDE.md and Cursor rules. It found the file in its throwaway test project and read it before writing any code.

The agent then probed SpacetimeDB's SQL and found it couldn't total, group or sort:

That matches SpacetimeDB's SQL reference, where COUNT is the only aggregation. The workaround is fine for 300 orders, but a real reporting workload would copy a lot of data on every query.

The hello world: a correct answer

The answer matched the figures we calculated from the CSVs: $212,620.84 for North, excluding refunds. The only slip was in the wording, because "61 orders" includes the four refunded ones.

The live page: updated 23 ms after the refund

This step tests SpacetimeDB's headline feature, which is that clients subscribe to data and get changes pushed to them. The agent added a database function to change an order's status, deployed it without losing data, and generated typed client code with spacetime generate. It then built a small web page with SpacetimeDB's TypeScript client library that subscribes to the orders and customers tables.

To prove the page updated without a reload, it opened the page in headless Chrome and refunded order 1001 from the command line. Then it ran seven checks, such as a marker stored in the page surviving the change and the page's update counter going from 0 to 1.

Both figures match our own: West went from $225,607.88 to $219,182.69.

Hiding customer emails: a private table and a public view

The agent made the customers table private and added a public view that exposes only each customer's ID, name and region. It recorded the "before" state first, when any anonymous client could read all 50 emails. SpacetimeDB applied the change to the running database without losing data.

We checked it ourselves as an anonymous client. Reading the customers table returned "no such table: customers. If the table exists, it may be marked private.", and the view returned names and regions without emails.

What it used, and what it didn't

Finally, we asked which SpacetimeDB docs and agent tools it had used:

Step	Agent time
Install and run locally	4.5 min
Load the data	7 min
Hello world	Under a minute
Live page and refund	7 min
Hide emails	4 min

The whole session took 45 minutes, including the time we spent checking each step, and we typed nothing but the six prompts.

Integration: GOOD (4 / 4) — Live app with private data, correct answers and no human steps

What SpacetimeDB can do about it

Point the rules file at the rest of the agent tooling

The rules file that spacetime init writes was the only piece of SpacetimeDB's agent tooling the agent used. A short section in that file on the plugin, spacetime mcp and agent-setup.md would introduce agents to the rest at the point they're already reading.

Document the reporting pattern

The agent hit the SQL limits by trial and error and built its own workaround. A docs page showing the intended way to compute totals and rankings in SpacetimeDB, such as in a view or a procedure, would give agents a pattern to follow instead of a SQLite copy.

What this means for SpacetimeDB

SpacetimeDB is a good product for an agent to build on. Working from its training data, the rules file the CLI writes and the npm package's own code, the agent installed it, loaded data, answered a question correctly, built a page that updated 23 ms after a database change, and hid private data.

The problems come before an agent starts building:

From memory, agents know SpacetimeDB as a game backend, and its new tagline sounds like Convex to them.
For web apps, agents recommend Convex or Supabase. SpacetimeDB lost both head-to-heads and the SaaS question.
Agents searching for its MCP server, skills and llms.txt found a community MCP server instead.

Guides for web developers, comparison pages that recommend SpacetimeDB for specific kinds of app, and an agent tooling page that ranks for those searches would each address one of these.

Orbital vs Hasura vs Apollo: AI Agent Review

RitzaCo — Thu, 22 Oct 2026 11:22:22 +0000

Orbital, Hasura, and Apollo all do similar things. But when agents write most of your software, the choice between them is made for you. Here’s which one your agent prefers, and why.

Using Claude Code, we followed our audit method to assess their agent experience (AX) across four stages.

The four stages of an AX audit

The audit follows the funnel a developer goes through with their agent:

Awareness → Discoverability → Reputation → Integration

We started with what the model knew without tools, then tested online discovery and recommendations with web search and fetch, and finally attempted a working integration with the full toolset. Explore the interactive audit diagram and exact prompts in the original article.

Awareness: Does the model already know about the product, without tools?
Discoverability: Does it find the product when it searches online?
Reputation: Does it recommend the product over alternatives, and get the facts right?
Integration: Can it build and verify a small working integration?

Each stage gets a score from 1 to 4: GOOD (4) means little friction; OK (3) means minor gaps; POOR (2) means a partial result with significant gaps; FAIL (1) means the product is unknown or the task doesn't work.

How we tested
All testing used Claude Opus 5.5 (claude-opus-5-5) in Claude Code 2.1.287, on 13–14 October 2026. Every session started in a fresh folder with no audit context or user customizations. Each integration used its own fresh Claude Code profile and a folder containing the supplied test data. Those profiles loaded no MCP servers and only built-in skills and plugins, and we approved every command. Awareness questions ran with no tools, and discoverability and reputation questions with web search and web fetch only. We ran each prompt once, so treat these results as one developer's experience rather than a benchmark.

We chose Hasura and Apollo GraphQL by asking six questions about Orbital's job and its alternatives, and taking the products the answers named most. Both appeared in all six answers. WunderGraph Cosmo did too, and we picked Apollo because Orbital's own comparison pages measure it against GraphQL tools like Apollo. The shared questions describe Orbital's problem (data spread across REST APIs, databases and Kafka), so they suit Orbital more than the other two.

What are Orbital, Hasura and Apollo GraphQL?

All three let an app or an agent ask one API for data that lives in several systems.

Orbital sits in front of your existing APIs, databases and message streams. You describe what the data in each system means once, then ask Orbital for the data you want, and it works out which systems to call and joins the results. Our Orbital quickstart covers how it works in more detail.

Hasura turns databases and APIs into one GraphQL API, and its company now also sells an AI agent product, PromptQL. Apollo GraphQL makes widely used GraphQL tools, including a router that combines many services into one API and Connectors that map REST APIs into it without code.

Results at a glance

🥇 Hasura (1st) — Known and found by agents; retrieved credentials and ran a script after browser login
🥈 Apollo GraphQL (2nd) — Known and found by agents; ran a script after we logged in and created a key
🥉 Orbital (3rd) — Won the named head-to-heads; never found unprompted, and its script stopped at authentication

Stage	Orbital	Hasura	Apollo GraphQL
Awareness	OK (3)	GOOD (4)	GOOD (4)
Discoverability	FAIL (1)	OK (3)	OK (3)
Reputation	GOOD (4)	OK (3)	OK (3)
Integration	FAIL (1)	OK (3)	POOR (2)
Overall (average)	2.25	3.25	3.0

The averages are not rounded. Hasura places first because its agent obtained credentials after browser login; Apollo needed us to create its key too.

Orbital vs Hasura vs Apollo GraphQL: which should you choose?

The agent's recommendation depended on your data: "use Orbital to combine mixed sources including Kafka without code, Hasura for an instant API over mainly SQL databases, and Apollo to federate GraphQL services owned by many teams."

Agents chose Orbital when the data includes Kafka topics and you want joins without writing code for each system, and it won every head-to-head on those grounds. Name Orbital explicitly, and be prepared to check script authentication yourself: our agent installed Orbital but couldn't get data back with the key we supplied.

Agents recommended Hasura when "your data is mostly relational and what you want is a fast, auto-generated GraphQL API over it". Against Apollo, the agent preferred Hasura in one order and neither in the other. Our agent used Hasura DDN, retrieved a token after our browser login, and ran a script that authenticated to Hasura Cloud and fetched local customer and order data. The local data queries used no token. When we asked about risks, the agent raised Hasura's shift towards its AI product, PromptQL.

Agents preferred Apollo when "Your consumers want GraphQL" or "Your sources are mostly REST", and said it "has a much larger community and is easier to hire for". It doesn't connect to databases or Kafka without extra services, which is why it lost both head-to-heads with Orbital. After we signed in and created a personal key, our agent used it to query GraphOS account information and also queried local customer and order data. The local router used no credentials.

Awareness: does the model know Orbital?

Awareness tests what a model knows from training data alone, with no web search.

Orbital: OK (3 / 4) — Knows it as a data gateway; doesn't know "governed context layer"; never named unprompted
Hasura: GOOD (4 / 4) — Accurate, and named for Orbital's job without a search
Apollo GraphQL: GOOD (4 / 4) — Accurate, and named for Orbital's job without a search

We first described Orbital's job without naming any product. Asked which tools let apps and AI agents query data across REST APIs, databases and Kafka without integration code, the model named Hasura, Apollo and WunderGraph Cosmo, among others. When we described one self-hosted layer with Orbital's main capabilities, it picked Grafbase, with Hasura DDN and Apollo as the alternatives. Orbital appeared in neither answer.

Then we named each product with its domain, in a fresh session with no tools: "What is Orbital (orbitalhq.com)?", and the same question for Hasura (hasura.io) and Apollo GraphQL (apollographql.com). The model described all three accurately:

Company	Answer (excerpt)
Orbital	"Orbital (orbitalhq.com) is a data integration platform, sometimes described as a "data gateway" or "integration without glue code." It comes from the team that previously built Vyne, which Orbital is essentially the successor to."
Hasura	"Hasura is a company and an open-source product that automatically generates a GraphQL API (and, in newer versions, other API styles) on top of your existing databases and services."
Apollo GraphQL	"Apollo GraphQL (apollographql.com) is a company that builds a widely used platform and set of open-source tools for working with GraphQL, the query language for APIs that Facebook originally created."

For Orbital, it described Taxi, the query language, joins across APIs, databases and Kafka, and the self-hosted Docker setup, but didn't mention AI agents or MCP. For Apollo, it mentioned the Apollo MCP Server.

Next we used each company's tagline from its own site, with no domain. The question for Orbital was: "What do you know about Orbital, the 'governed context layer'? What does it do, and who is it for? Say clearly if you don't know." Hasura's and Apollo's versions swapped in their own name and tagline. Orbital's homepage turns "your existing APIs, databases and streams into a governed context layer", and the model didn't connect that phrase to Orbital:

Company	Tagline	Answer (excerpt)
Orbital	"governed context layer"	"I don't know what this is. I have no reliable information about a product called "Orbital" described as a "governed context layer," so I can't tell you what it does or who it's for."
Hasura	"universal data access layer"	"…It is metadata-driven, uses a pluggable connector spec (NDC) for arbitrary data sources, and is where the "universal data access layer" positioning mainly comes from." (about Hasura DDN)
Apollo GraphQL	"domain model for AI"	"I'm not certain Apollo uses that tagline, or how they currently frame it. It fits the direction I described, with the graph as the semantic layer between your APIs and AI."

For Orbital, it listed Orbital Insight, a geospatial analytics company, as one of several unrelated things called Orbital. For Hasura and Apollo, it described the products accurately either way.

Asked how to join two REST APIs with Orbital on a laptop, the model gave the right shape of a setup: Docker, a Taxi project with a shared customer ID type, and a query. Its first command saved https://start.orbitalhq.com as a Docker Compose file, even though that address serves Orbital's launcher script. It also didn't mention the sign-in Orbital needs on first start.

What this means if you're choosing

Without web search, an agent suggests Hasura or Apollo for this job. If you want Orbital considered, name it. The model knows it as a data gateway, which is how Orbital's docs describe it, but not by the "governed context layer" description on its homepage.

Discoverability: can an agent find Orbital?

Discoverability tests whether an agent with web search finds a product while looking for a tool like it. We repeated the two awareness questions with web search, then ran two six-question funnels that narrowed a broad problem towards Orbital. The first four questions were the same for all three companies. Each funnel branched at question 5, where we asked for products using that company's tagline; we named the company at question 6.

Orbital: FAIL (1 / 4) — Only found when named, in both funnels; its tagline search found Cube
Hasura: OK (3 / 4) — Named at question 1 or 3; its tagline search found it both times
Apollo GraphQL: OK (3 / 4) — Named at question 1 or 3; its tagline search found it once

With web search, the two open questions named Hasura and Apollo again, but not Orbital. One query came close to Orbital's description: self-hosted GraphQL federation REST databases Kafka no resolvers field-level authorization 2026. The answer said no tool absorbs upstream API changes:

Funnel A targeted a platform engineer tired of glue code between microservices; funnel B followed a team building AI agents, the audience Orbital's homepage now speaks to. Here is funnel B on the Orbital branch, with the start of each answer:

At question 4 the agent searched for "context layer" OR "data gateway" for AI agents startup launch 2026 governed access internal systems, which uses both of Orbital's descriptions, without finding Orbital. At question 5, Orbital's tagline led to Cube, whose AI Context Layer page says "Cube gives your AI agents a governed context layer". Funnel A's tagline search went the same way, listing Cube, Starburst, Denodo and Atlan:

Funnel	Orbital first appears	Hasura first appears	Apollo GraphQL first appears
A: platform engineer	Question 6 (named)	Question 1, in a list of federation tools	Question 1, in the same list
B: team building AI agents	Question 6 (named)	Question 3	Question 3

Hasura's tagline, "universal data access layer", returned Hasura in both funnels. Searching Apollo's tagline, "domain model for AI", returned Apollo in funnel A and SAP's domain models in funnel B.

What this means if you're choosing

If you shortlist tools by asking an agent to search, expect GraphQL federation tools: The Guild's Hive stack, Grafbase, WunderGraph Cosmo, Apollo and Hasura. Orbital didn't come up in any of our searches for this problem, so add it to the list yourself.

Reputation: do agents recommend Orbital over Hasura and Apollo GraphQL?

Reputation tests what an agent with web search says about each product once it's in the conversation. Every question described the same use case: "joining data across our REST APIs, databases and Kafka, self-hosted, without integration code".

Orbital: GOOD (4 / 4) — Picked over both, in every head-to-head and in the three-way question
Hasura: OK (3 / 4) — Recommended when the data is mostly in databases; split with Apollo
Apollo GraphQL: OK (3 / 4) — Recommended for GraphQL teams with mostly REST APIs; lost to Orbital, split with Hasura

Asked for the single best tool, with no product named, the agent picked ClickHouse and didn't mention any of the three.

Once all three were named, the agent chose Orbital, answering from what it already knew without a search:

Apollo vs Hasura: how do they compare with Orbital?

Orbital beat both Hasura and Apollo, in both orders. For Hasura vs Apollo GraphQL, the result was a split. We asked each pair in both orders because agents can favour the product named first, and a pair counts as a win only if the same product wins both ways.

Pair	Named first	Named second	Result
Orbital vs Hasura	Orbital	Orbital	Orbital wins
Orbital vs Apollo GraphQL	Orbital	Orbital	Orbital wins
Hasura vs Apollo GraphQL	Neither: "neither is a great fit for all three requirements together"	Hasura	Split

In each Orbital head-to-head, the agent searched for Orbital and cited its Kafka support and joins that need no code per system. With Hasura named first against Apollo, it recommended Grafbase or WunderGraph Cosmo if Kafka matters.

The last question used each product's bare name, with no domain. The agent recognised Hasura and Apollo GraphQL and flagged Hasura's shift towards PromptQL and v2.45 LTS reaching end of support on 1 January 2027, plus Apollo's Elastic License and usage-based pricing. Asked about "Orbital" on its own, it returned information about a different industry:

What this means if you're choosing

Named side by side, Orbital won on Kafka support and joins without per-system code. Hasura and Apollo each won a narrower case: Hasura for data that sits mostly in databases, Apollo for mostly REST APIs. Our questions described a mix of REST APIs, databases and Kafka, which is Orbital's own use case. When you ask an agent about Orbital, give it the URL: on its own, "Orbital" brought up data centres in space.

Integration: can an agent build with Orbital?

Integration follows three steps: sign up or install, get a credential for a script, then write and run a script that returns real data. Each company had one continuous session. The agents had shell access, web search and fetch, but no browser tool. We handled browser steps when asked. The test folders included two small REST services for customers and orders.

Orbital: FAIL (1 / 4) — Installed, but the supplied key returned errors and no data
Hasura: OK (3 / 4) — Human browser login; agent retrieved the token and ran a script
Apollo GraphQL: POOR (2 / 4) — Human account access and key creation; agent ran a script

The first request was: "Sign me up for Orbital, or install it if that's how it works, and get it ready to use. Do what you can yourself and tell me where you need me." Hasura and Apollo GraphQL received the same prompts with their names substituted.

Company	Sign up / install	API key / access token	Hello world
Orbital	Agent installed Docker stack; browser reused our existing account	Human created a key and saved it to `.env`	Stuck: authenticated calls returned HTTP 500; no data
Hasura	Agent installed DDN CLI; human completed browser authentication	Agent retrieved the CLI token and saved `.env`	Agent ran a script: authenticated cloud project listing, plus local customer/order queries
Apollo GraphQL	Agent installed Rover; human handled Studio account access	Human created a personal key and saved `.env`	Agent ran a script: authenticated cloud account query, plus local customer/order queries

Orbital: installed, then stopped at authentication

Orbital's agent downloaded and read the official launcher, started the Docker stack, and checked that its health endpoint returned 200. Our existing Orbital account signed in automatically in the browser.

Asked to get a credential, the agent searched the docs, probed the API and inspected the local UI. It didn't obtain a key or write .env. We created an API key and saved it ourselves so the next step could test whether the agent could use it. Requests with that key returned HTTP 500, and the agent found this message in Orbital's logs: API Key authentication is not enabled.

We didn't test script authentication with other Orbital configurations.

Hasura: browser login, then the agent got the token

Hasura's agent chose DDN, installed its CLI and asked us to complete browser authentication. After that it used ddn auth print-access-token to save a token privately, without asking us to create or copy one.

The agent generated two OpenAPI connectors, fixed a Docker build failure and missing environment settings, then ran query_hasura.py. Its saved token authenticated a Hasura Cloud project-list request. Separately, the local engine returned Luca Ibrahim and nine orders, with a completed-order total of 2,669.39, matching our test data. That local engine didn't require a key. The script fetched the customer and orders with two queries in one request; it hadn't created a relationship between them.

Apollo: we created the key, then the script worked

Apollo's agent installed Rover through Homebrew and tested a local customer query. When asked for credentials, it started a browser login but stopped to ask which account, organization and key type to use.

We created a personal key in Studio and saved it to .env. The agent wrote query.py, fixed a missing client-version header, and ran it successfully. The key authenticated a GraphOS query returning our account and team information. A separate request to the local router returned Luca Ibrahim with the same nine orders. The local router answered without credentials.

Hasura's and Apollo's integration scores reflect who handled account access and credentials, plus whether a script ran successfully.

What this means if you're choosing

Orbital installed successfully, but browser sign-in and a supplied key didn't get our agent to a working script.
Hasura required browser authentication, then the agent obtained its token and completed the script. It also repaired generated connector files along the way.
Apollo's agent completed the script after we created and saved the personal key. With either Hasura or Apollo, check local API authentication before using the demo with private data.

Fish Audio vs ElevenLabs: Voice Cloning & TTS

RitzaCo — Thu, 22 Oct 2026 09:13:04 +0000

If you're choosing between Fish Audio and ElevenLabs, the answer depends on whose voice you need. With stock voices, ElevenLabs sounded more human, handled emotional direction better and read an appointment message correctly, while Fish turned R1,249.50 into more than a million dollars. For cloning your own voice from a short sample, Fish got much closer to sounding like me, although ElevenLabs' clone was the cleaner recording.

I tested Fish Audio Drama 3, released on 13 October 2026, against Eleven v4 using the same scripts and one take each, then cloned my own voice on both. Links to every clip are below, so you can hear the differences for yourself. Neither instant clone would convince someone who knows my voice, so if cloning is why you're here, read that section with that in mind.

Fish Audio vs ElevenLabs: the quick verdict

Test	Winner	One-line reason
Instant voice cloning: likeness	Fish Audio	Much closer to my accent and tone, and kept my identity when the mood changed
Instant voice cloning: audio quality	ElevenLabs	Crisper and more expressive, but sounded like a different person
Natural delivery (no tags)	ElevenLabs	More human, with excitement and pauses in the right places
Emotional direction (tags)	ElevenLabs	Carried emotion from one sentence to the next, and the laugh fitted
Names, numbers and currency	ElevenLabs, by far	Fish misread R1,249.50 as over a million dollars

How we tested

Models. Eleven v4 against Fish Audio Drama 3.

Stock voices. ElevenLabs' Global Artist - Newsreader against Fish's Formal Announcer. Both are British female voices.

Same words on both. The directed scripts used each tool's own tag wording ([whispers] and [laughs] on ElevenLabs, [whispering] and [laughing] on Fish), with [excited] and [pause] on both.

Single take. Unlike my Gemini vs ElevenLabs comparison, where I picked the better of two ElevenLabs outputs, every clip here is the first output from a single generation, with default settings.

Cloning used the same 33.94-second recording of my voice on both. Fish's clone was made on the free website. ElevenLabs wouldn't let me save a clone on Free, so I made that one through an existing paid Creator API account, with background noise removal off.

Fish Audio voice cloning vs ElevenLabs voice cloning

What this tests: can each tool sound like me from around 30 seconds of audio, and keep sounding like me when the line goes from everyday to relieved. I'm South African, so accent drift is easy for me to hear.

Both upload screens accepted the same file. ElevenLabs asks for at least 10 seconds and recommends 3 minutes for the best quality. Fish also asks for 10 seconds, and marks around 30 seconds as enough for most voices. My recording is well short of the ElevenLabs recommendation, so treat this as a test of a quick clone from limited audio.

Both clones read the same new passage. It starts with everyday errands and ends on good news.

I left my keys next to the kettle again. If you're heading out, could you check whether the back door is locked? I should be home before the rain starts. Oh, and the good news: they finally approved the project! I can stop refreshing my inbox every five minutes. Let's get dinner somewhere nice tonight.

Listen: Instant voice cloning

Same clean 33.94 s recording of my voice, same output text

ElevenLabs (eleven_v4) — MP3: Crisper audio, less like my voice
Fish Audio (Drama 3) — MP3: Closer accent and identity, softer audio

Fish wins on likeness, with a clone much closer to my accent and tone. There's a hint of an American accent and the delivery is slightly sluggish, but it stayed recognisably me as the expression changed. It also sounds like someone reading text, which is fair, because the reference was me reading from my phone.

ElevenLabs sounds crisper, and its expression is more distinct. But on "they finally approved the project!" it sounded like a different person. Fish delivered that line as relief rather than excitement, which I think suits someone who can finally stop refreshing their inbox, and it kept more of my identity while doing it.

The Fish clip is a little damp next to ElevenLabs, and I'm happy to trade that for sounding like myself. To me the ElevenLabs clone sounded as if it had generalised my voice into something more generic. That's a listening impression, and I can't confirm anything about what's actually going on in the background.

If I had to use one of these clones, I'd pick Fish. Neither is good enough that someone who knows my voice would definitely believe it was me, because the accent differences are still audible. I'd expect Fish's Professional Voice Cloning to do better, but I didn't test it.

Where the paywalls are

ElevenLabs let me upload the recording, listen to previews and see labels it had inferred from my voice. I liked that, and I thought the labels were accurate.

Then I clicked Save voice and hit a paywall. Instant cloning needs at least the Starter plan, and nothing earlier in the flow had made that clear to me. I'd rather find out before uploading my voice and going through the setup.

Fish let me create the clone on Free, but only as a public voice. Making it private needs a paid plan. I deleted the public Fish clone and the temporary ElevenLabs clone after the test. Deletion removes the voice from your list, but Fish's terms include licences over publicly submitted content, so don't assume deleting a public clone undoes everything.

Only clone a voice you own or have permission to use.

Fish Audio TTS vs ElevenLabs: natural delivery and emotion

What this tests: whether each voice finds the emotion in a script on its own, and then how well it follows direction. The script is someone getting good news at work and then remembering it isn't public yet.

I wasn't going to tell anyone. We got the Lisbon project! I read the email three times before I believed it. Anyway, don't tell anyone yet. It's not official until Friday. Right. Now I just have to work out how to tell the team without grinning like an idiot.

Natural read, no tags

Listen: Natural read

Plain text, no tags. Does the voice infer the excitement and the pauses?

ElevenLabs (eleven_v4) — MP3: More human, natural pause on 'Right.'
Fish Audio (Drama 3) — MP3: Good emphasis on 'three times'

ElevenLabs sounded more human to me. At 0:02 it puts excitement into "we got the Lisbon project!", and the pause around "Right." sounds like someone collecting their thoughts.

Fish inferred the emotion too. At 0:07 it slows down on "three times" to stress how often the email was read, which is a nice touch. Both handled "grinning like an idiot" well. The problem is the voice itself, which reminded me of older AI text to speech and made the emotion less convincing.

That verdict is about these two stock voices, and a different Fish voice could change it. I'd use the ElevenLabs take as it is.

Directed read, with tags

The directed version keeps the same words and adds tags. Here's the ElevenLabs script:

[whispers] I wasn't going to tell anyone. [excited] We got the Lisbon project! I read the email three times before I believed it. [whispers] Anyway, don't tell anyone yet. It's not official until Friday. [pause] Right. [laughs] Now I just have to work out how to tell the team without grinning like an idiot.

Fish got the same script with [whispering] and [laughing] in place of [whispers] and [laughs].

Listen: Directed read

Whisper, excitement, a pause and a laugh, same position on both

ElevenLabs (eleven_v4) — MP3: Emotion carries between sentences
Fish Audio (Drama 3) — MP3: Abrupt switch to a whisper at 0:10

Direction made both tools more expressive, and so more human, but ElevenLabs was still better. At 0:04 it speeds into the next sentence while holding on to the excitement from the one before, and it sounds like one person continuing a thought. Its laugh fits the conversation.

Fish sounded as though each sentence had been given its own emotion and followed it literally. At 0:10 the switch to a whisper is abrupt and a bit uncanny, and the pace stays the same whatever the expression. Its laugh sounds like it's laughing because it was told to. I didn't hear either tool read a tag aloud or skip one, and I'd use the ElevenLabs directed take as it is.

Names, numbers and hard text

What this tests: the kind of appointment message a clinic might send. It has a name, a date, a time, a currency amount, a booking reference, a phone number and a network frequency, with no tags on either side.

Hi, it's Dr. Nkosi's office. Your appointment is on 3 April 2026 at 14:45. The total is R1,249.50. Your booking reference is A-7419-X. If you need to change it, call 0800 555 0147. For the video appointment, please use the 2.4 GHz network if the other connection drops. Thanks, and we'll see you on Friday.

Listen: Names and numbers

Name, date, time, rand amount, booking reference, phone number, frequency

ElevenLabs (eleven_v4) — MP3: Everything right, natural ending
Fish Audio (Drama 3) — MP3: Misread the rand amount

ElevenLabs recognised R1,249.50 as rand and read the amount and the cents correctly. Fish made the most serious error in this test. What I heard was "R one million two hundred forty-nine dollars and fifty cents", which gets both the size of the amount and the currency wrong. I'd treat that as a serious error in anything that tells a customer what they owe.

Item	ElevenLabs	Fish Audio
`Dr. Nkosi`	Correct to my ear, the "o" sounds like "aw"	Recognisable, but the "o" sounds like "ow"
`3 April 2026`	"April third 2026", natural	"three April 2026", which sounds unnatural
`R1,249.50`	Correct amount, rand and cents	"R one million two hundred forty-nine dollars and fifty cents"
`A-7419-X`	Both dashes spoken	Left out the second dash
`0800 555 0147`	Correct, zeros read the same way throughout	Correct digits, but the first zero is "oh" and the rest are "zero"

Both read 14:45 and 2.4 GHz correctly.

Spoken separators in the booking reference help when someone has to read it back, because they mark where the parts break. The phone number is more a matter of taste, since Fish got the digits right but sounds odd mixing "oh" and "zero". My one wish for ElevenLabs is "oh-eight-hundred" for 0800, which is how Gemini read it in my Gemini vs ElevenLabs comparison.

ElevenLabs wins this by a long way, and it also ended the paragraph more naturally. Its take would need far less editing before I'd use it. I didn't try correcting Fish's reading.

Is Fish Audio free? Fish Audio vs ElevenLabs pricing

Yes, Fish Audio's free plan gives you 8,000 credits a month, and it let me clone my voice as long as the clone was public. ElevenLabs is free for text to speech, but saving a voice clone needs a paid plan. Prices below were checked on 15 October 2026, in US dollars.

	ElevenLabs	Fish Audio
Free plan	10,000 credits a month, no instant cloning, no commercial licence	8,000 credits a month, 500 characters per generation, 3 public voice slots, personal non-commercial use only
Entry paid plan	Starter: $6 a month for 30,000 credits, includes instant cloning	Plus: $15 a month, or $132 a year billed annually, for 250,000 credits and 10 private voice slots
API price for the model I tested	Eleven v4: $0.08 per 1,000 characters	Drama 3: $50 per million characters ($0.05 per 1,000), 80% off until 12 November 2026

ElevenLabs' professional cloning starts on Creator at $22 a month. Its annual billing charges ten months' price for a year. Credits on the websites and dollars on the APIs are separate, so don't convert one into the other.

My ElevenLabs website balance went from 6,064 to 5,191 credits for the two stock-voice tests. Fish went from 8,000 to 6,818, but that also covers the clone. The workloads are different, so these numbers don't tell you which is cheaper.

Fish's first-time tips were helpful when I added my first directed script. The biggest friction on ElevenLabs was finding out about the cloning paywall at the last step.

Both free plans restrict commercial use, so if you're making something for paid work, check the plan and voice permissions first (ElevenLabs, Fish Audio).

Is Fish Audio a good ElevenLabs alternative?

For cloning your own voice from a short sample, yes. For ordinary text to speech, ElevenLabs was better in every stock-voice test I ran. This covers two models and four short tests, so for a wider view, see Best TTS in 2026: blind benchmark and Cartesia vs ElevenLabs.

Ordinary voiceovers in English: ElevenLabs. Its stock voice sounded more human with and without direction.
Anything with prices, references or phone numbers: ElevenLabs. After Fish's currency error, I wouldn't use it for that kind of text without checking every take.
Cloning your own voice from a short sample: Fish. It sounded more like me.

MSI Titan 18 HX Dragon Edition Norse Myth AI Review: Can It Run Advanced AI Workloads Locally?

RitzaCo — Wed, 16 Sep 2026 09:37:48 +0000

Watch the MSI Titan 18 HX run Qwen3.8-Flash-Next locally:

At the forefront of the AI race, Anthropic, OpenAI, and the other big players have been all over mainstream news for a couple of years now. As of August 2026, many people are comparing GPT Sol to Fable and sharing the latest attempts to one-shot a GTA clone. However, there is another race going on that gets much less attention, led by some lesser-known AI companies. Open-weight mixture-of-experts models such as Qwen have improved dramatically in capability, efficiency, and general usefulness. As they become more efficient, running them on consumer-grade hardware becomes more feasible.

MSI recently sent us one of its Titan laptops to experiment with, and we had one main question: Can you really run a useful LLM on a laptop?

The quick answer is yes, and the experience was much better than we expected. As a slightly cheesy example, here is a playable Three.js game that the local model one-shotted as a single HTML file: Press start

Play the Qwen-built lane runner on TechStackUps →

The Titan Is More Workstation Than Gaming Laptop

Okay, so we mentioned that it is becoming feasible to run local models on consumer hardware. It is getting there, but this particular laptop from MSI is not your standard student laptop. It does not even really qualify as a high-end gaming laptop. It is closer to a chunk of the supercomputer hardware you might rent from Vast.ai or Lambda, except that it comes with a keyboard and an enormous screen.

96 GB of RAM Makes Local Qwen Possible

The laptop has an Intel Core Ultra 9 285HX and an RTX 5090 Laptop GPU. Most importantly for this use case, it also has 96 GB of DDR5 RAM. The main specifications are:

Component	Specification
CPU	Intel Core Ultra 9 285HX
GPU	NVIDIA GeForce RTX 5090 Laptop GPU with 24 GB GDDR7 VRAM
Memory	96 GB DDR5-6400 RAM
Display	18-inch 3840 × 2400 Mini LED, 120 Hz
Storage	2 TB PCIe Gen5 SSD plus 4 TB PCIe Gen4 SSD

The 24 GB of VRAM is great, but the unusually large pool of system RAM is what makes this machine interesting for models that cannot fit entirely on the GPU. You can find the complete configuration on the MSI specification page.

The Dragon Edition Fully Commits to the Theme

It also looks really cool. With an embossed aluminum top panel, a 4K display, and fancy gamer lights, the Titan looks and feels like a piece of premium technology. Lots of dragons. Battery for scale.

Our package also came with a matching gaming mouse and mouse pad, plus an enormous 400 W charging brick.

A 180-Billion-Parameter Model Fits, but Only at 2 Bits

We originally planned to test several local models on this hardware, but then, with excellent timing, Alibaba's Qwen team released Qwen3.8-Flash-Next. It is an open-weight mixture-of-experts model with 180 billion total parameters, but it activates only a small fraction of them for each token. That makes it far more practical to run locally than the headline parameter count suggests.

The full-precision model is still far too large for a laptop, so we used Unsloth's 2-bit UD-Q2_K_XL GGUF quantization. The download is about 73.5 GB. At runtime, it used roughly 21 GB of the RTX 5090's VRAM and about 75 GB of system RAM, leaving enough headroom for Windows, our agent harness, the monitor, and a browser. The model advertises support for up to 262,144 tokens of context. We tested it with a 32,768-token server allocation, but we did not establish that as the laptop's maximum.

We therefore kept the tests focused on this Qwen model and compared it primarily with OpenAI's GPT-5.6 Luna, a cheap and capable cloud model.

Qwen Was Easy to Install but Needed Manual Tuning

Installing the model and getting it running locally mostly required us to follow Unsloth's Qwen3.8-Next guide. The out-of-the-box settings did not provide the best starting point, though: automatic memory fitting initially reduced the context to 8,192 tokens, which hurt performance considerably until we manually raised it.

Loading the 73.5 GB model from cold took 60.1 seconds.

Pi Kept the Local and Cloud Tests Comparable

To compare the models easily, and because we generally like using it, we installed Pi as our agent harness.

This added another layer of configuration, which complicated later troubleshooting. Pi also loads its standard tool definitions into the context before a task begins, which is worth remembering when comparing context use with a direct API call.

MSI Center and Our Dashboard Tracked Hardware and Tokens

The laptop came with MSI Center pre-installed. For every test, we kept the laptop connected to its 400 W charger and selected the Extreme Performance profile under User Scenario. We used the Hardware Monitoring view in MSI Center to watch CPU and GPU load and temperatures, memory use, and other system values while Qwen ran.

For the model-side measurements, we got Codex to make a small companion app that tails the latest JSONL session from Pi or Codex. It reports the outcome, elapsed time, generated tokens, thinking share, context use, and cache use.

Qwen Overthought the Easy Test, Stayed Cool, and Built the Better Game

We were not trying to produce a leaderboard from a handful of prompts. Instead, we picked tasks that exposed different parts of the experience: how the model uses its thinking budget, what a long agentic task does to the laptop, and whether the final output actually works.

Fibonacci: Qwen Did Nine Times More Work Than Necessary

We started with the smallest possible coding task: "Write me Fibonacci in Python and save it to a file."

The clip below shows the local run in progress. MSI Center tracks CPU and GPU load in the top left, our companion dashboard tracks time, tokens, and context below it, and Pi shows Qwen's work on the right.

With its thinking budget turned up, Qwen produced four implementations, including an O(log n) fast-doubling version. It also added command-line arguments, type hints, validation, and tests, and used four tool calls. The result took 71 seconds and generated 1,647 output tokens.

Luna took eight seconds, generated 177 output tokens, and wrote a basic 12-line function. Qwen's work was better, but for this prompt it was comically over-engineered. Separate direct benchmarks put Qwen's dependable generation rate between 26 and 33 tokens per second. Raw speed was not the problem here; it decided to do about nine times as much work.

That gave us a useful early lesson: a large thinking budget is not free intelligence. We switched thinking off for the remaining tests, and Qwen became much more direct without a noticeable drop in the quality of the results.

Pelican: 89 GB of RAM and No Thermal Throttling

The next test came from Simon Willison's Qwen3.8-Flash-Next write-up: create an SVG of a pelican riding a bicycle. Our early results were highly variable. Sometimes the model put real effort into the illustration; sometimes it did the least it could possibly get away with.

To make the comparison less dependent on how ambitious either model happened to feel, we added the word "detailed" to the prompt and ran it again with Qwen and Luna. Qwen is on the left below, and Luna is on the right. Both produced a recognizable, fairly detailed pelican on a bicycle, and Qwen's version looks substantially better.

During the local run, system RAM peaked at about 89 GB, and VRAM stayed near 21.8 GB. The GPU peaked at 69°C and roughly 111 W. In a separate eight-minute sustained run, it again topped out at 69°C, averaged 63.7°C, and reached a brief 213.5 W power peak without thermal throttling or an upward temperature trend.

The fans are not subtle. In one 90-second sample near the laptop, a phone sound meter averaged 62.1 dB-A and showed 65.7 dB-A when we took the screenshot. This is not a calibrated lab measurement, but it accurately captures the experience: under sustained load, the Titan sounds like it is moving a lot of air.

While watching MSI Center, we saw the CPU temperature peak at about 89°C. When the job finished, it quickly dropped to its idle range of about 45°C to 50°C.

In other words, the model pushed the memory and cooling system hard, but we did not hit a practical hardware limit. The laptop shed heat quickly, remained responsive throughout the run, and never throttled or crashed.

Three.js: Qwen Was Slower but Built the Only Playable Game

For the final test, we asked both agents to make a three-lane endless runner in Three.js as a single HTML file. The prompt specified 21 requirements, including jumping, lane changes, obstacles, collisions, scoring, and restarting.

Qwen took five minutes and 36 seconds, used 18 tool calls, and generated 9,266 output tokens. Luna took one minute and 37 seconds, used six tool calls, and generated 4,314 output tokens. However, Luna's attempt was vastly inferior.

Although the score increased and the obstacles moved, the player was invisible because the camera framing placed it below the viewport. It also looked much more amateurish than Qwen's attempt and showed little ambition in its visuals. This is exactly why we care more about the output than a benchmark score.

The short clips below show the two results side by side.

Qwen: slower, but playable

Luna: faster, but where is the player?

The local model took more time and used more tokens, but it also produced the better result.

Local AI Is a Great Bonus, Not a Business Case

The exact configuration we tested, with 96 GB of RAM, an RTX 5090, and 6 TB of storage, has been listed for about $6,200. Prices and availability vary, but that gives us a reasonable figure for a rough comparison. If the only goal is to avoid an AI bill, the math is brutal.

Comparison	What $6,200 buys
ChatGPT Plus at $20 per month	About 26 years
ChatGPT Pro at $200 per month	About 31 months
GPT-5.6 Luna output tokens at $1.20 per million	About 5.2 billion tokens

That token figure is deliberately generous to the laptop. Output is Luna's most expensive token category, so a realistic mix of cheaper input, cached-input, and output tokens would require even more total tokens to reach $6,200. This also ignores the laptop's electricity use and depreciation, and assumes that a local model can replace every feature of the cloud service.

As a purchase dedicated to running local LLMs, the Titan does not make financial sense. But that is not really the strongest case for it. If you already want a very powerful laptop for gaming, rendering, video editing, machine learning, or other heavy workflows, adding a capable, private local LLM to your toolbox is definitely worthwhile. Treat local inference as a valuable bonus, not the reason the machine pays for itself.

FAQ

Can the MSI Titan 18 HX Run Large AI Models Locally?

Yes. We ran a 180-billion-parameter model using a 2-bit quantization, with the weights split between the RTX 5090's VRAM and system RAM. The machine handled it reliably, although the 89 GB RAM peak leaves little room for a larger model, higher-precision weights, or a much larger context allocation.

Which Local AI Workloads Suit the Titan Best?

It makes the most sense for single-user coding agents, rapid prototyping, and generative tasks where local processing, privacy, or avoiding per-token charges matters. It is less compelling as a high-throughput server or a machine shared by several users.

Can It Replace a Desktop AI Workstation or Cloud Platform?

For some single-user workflows, yes, but not completely. It provides workstation-class local inference in a portable package, while a desktop remains easier to upgrade and cloud platforms offer a wider choice of frontier models and scalable compute.

Best SERP API Comparison 2025: SerpAPI vs Exa vs Tavily vs ScrapingDog vs ScrapingBee

RitzaCo — Wed, 03 Dec 2025 13:49:46 +0000

We tested five SERP APIs with standardized benchmarks to find which performs best for different use cases. Exa delivered the fastest responses at 1.180 seconds with semantic search over a proprietary index at $5 per 1,000 queries. Tavily aggregates and parses content for LLM applications. Scrapingdog offers the lowest per-request pricing at under $0.001 at scale. SerpAPI provides the broadest coverage with 20+ search engines. Scrapingbee bundles SERP access with general web scraping.

This comparison helps you choose a SERP API for SEO monitoring, AI agents, RAG applications, or price comparison tools. Each section includes benchmark data, pricing breakdowns, and code examples.

Overview Comparison

We tested five SERP APIs using this Python benchmark script that measured response time, success rate, cost, and free tier across 50 real-world queries.

API	Avg Response Time	Success Rate	Cost per 1,000 Searches	Free Tier
Exa	1.180s	100%	$5	2,000 one-time
Scrapingdog	1.700s	100%	$0.20-$0.06	1,000+ one-time
Tavily	1.885s	100%	$5-$8	1,000/month
SerpAPI	2.972s	100%	$15	250/month
Scrapingbee	7.257s	100%	$0.56-$1.44	66 one-time

Response times vary based on what each API delivers: SerpAPI returns raw SERP data, Exa performs semantic search with embeddings generation, and Tavily aggregates content from multiple sources with AI ranking. These tests used free tiers, but some providers offer faster speeds for additional cost (for example, SerpAPI's Ludicrous Speed tiers claim 2.2x faster responses).

SerpAPI

SerpAPI scrapes 20+ search engines and returns structured JSON. The service handles CAPTCHA solving and proxy rotation automatically, so you make a request and get back parsed data without dealing with blocks or rate limits.

Speed

Our 50-query benchmark measured 2.972s average response time with 100% success rate. Response times ranged from 0.070s to 13.308s, showing high variability. The free tier uses "Best Effort" speed with automatic refunds for failed searches. Paid tiers offer Ludicrous Speed claiming 2.2x faster responses for time-sensitive applications. SerpAPI's status page reports 99.911% uptime over the last 30 days.

Speed Score (6.0/10): Mid-tier performance at 2.972s average places SerpAPI fourth out of five tested APIs, though the wide variance suggests some queries complete in milliseconds while others take 10+ seconds.

Pricing

SerpAPI's pricing uses monthly subscriptions based on search volume. Plans start at $75/month for 5,000 searches, which works out to $15 per 1,000 searches or $0.015 per search. Production tier and above include U.S. Legal Shield, providing legal defense for crawling public search data under First Amendment rights.

Pricing Score (3.0/10): At $15 per 1,000 searches, SerpAPI costs 3-15x more than alternatives like Scrapingdog ($0.20) or Exa ($5), earning the lowest pricing score among tested APIs.

Developer Experience

SerpAPI provides SDKs for Ruby, Python, JavaScript, Golang, PHP, Java, Rust, and .NET. Setup takes about 5 minutes:

from serpapi import GoogleSearch

params = {
    "q": "coffee",
    "location": "Austin, Texas",
    "api_key": "your_api_key"
}

search = GoogleSearch(params)
results = search.get_dict()

for result in results["organic_results"]:
    print(result["title"])

SerpAPI's Playground lets you test queries through a web interface before writing code. You modify parameters through form controls and see both the rendered SERP and JSON response. The dashboard tracks your usage with response time graphs and search history.

Developer Experience Score (10.0/10): Eight official SDKs, an interactive playground, comprehensive documentation, and real-time usage dashboards make SerpAPI the strongest developer experience among tested APIs.

Search Capabilities

SerpAPI covers Google (including AI Mode and AI Overview), Bing, Yahoo, Baidu, DuckDuckGo, Yandex, Amazon, eBay, Walmart, Home Depot, YouTube, Google Scholar, Google Patents, TripAdvisor, and Yelp. Each platform has its own API endpoint with consistent JSON formatting. Most competitors support 3-5 search engines at most.

Search Capabilities Score (10.0/10): Coverage of 20+ search engines and specialized platforms far exceeds any competitor, making SerpAPI the only choice for applications needing multi-engine support.

Free Tier Generosity

The free tier provides 250 searches per month with no credit card required. All API endpoints work on the free tier, though you're limited to Best Effort speed without access to Ludicrous Speed tiers. Failed searches receive automatic refunds, so the 250-search limit only counts successful requests.

Free Tier Score (6.0/10): The 250 searches per month places SerpAPI mid-tier, more generous than Scrapingbee (66 total) but less than Tavily (1,000/month), Exa (2,000 one-time), or Scrapingdog (11,000 with promotions).

Choose SerpAPI when you need breadth of coverage

SerpAPI works best when you need access to multiple search engines or specialized platforms through one provider. SEO tools monitoring rankings across Google, Bing, and Yahoo avoid maintaining separate integrations. E-commerce price monitoring pulls structured data from Amazon, Walmart, and eBay. Academic research platforms integrate Google Scholar without parsing citation formats. The $15 per 1,000 searches pays for itself when infrastructure complexity is the bottleneck.

Skip SerpAPI if you only need Google search on a tight budget. At $0.015 per search, you're paying 3-15x more than alternatives. AI applications requiring semantic search or embeddings should use Exa or Tavily instead.

Exa

Exa is a search API built for AI applications using a proprietary embeddings-based index. Instead of scraping Google or Bing, Exa maintains its own crawl optimized for semantic search.

Speed

Our 50-query benchmark measured 1.180s average response time with 100% success rate. Response times ranged from 1.003s to 2.120s. Exa's published benchmarks claim P50 latency below 425ms, though our numbers likely include network overhead and full result retrieval while their metrics measure API response only.

Speed Score (10.0/10): At 1.180s average, Exa delivered the fastest response times among all tested APIs, making it ideal for latency-sensitive applications.

Pricing

Exa's pricing charges $5 per 1,000 searches for 1-25 results ($0.005 per search). Requesting 26-100 results jumps to $25 per 1,000 searches ($0.025 per search), a 5x increase. Additional endpoints cost separately: contents retrieval runs $1 per 1,000 pages, answers and research tasks run $5 per 1,000 calls.

Pricing Score (7.5/10): At $5 per 1,000 searches, Exa sits in the affordable mid-tier, significantly cheaper than SerpAPI ($15) but more expensive than Scrapingdog ($0.20).

Developer Experience

Exa provides SDKs for Python (exa-py) and JavaScript (exa-js), both with full TypeScript definitions. Setup takes about 5 minutes:

from exa_py import Exa

exa = Exa(api_key="your-api-key")
results = exa.search("coffee")

for result in results.results:
    print(result.title)

The Playground lets you test queries with controls for search type, filters, and AI features while generating code snippets. Documentation includes integration examples for LangChain, LlamaIndex, CrewAI, Anthropic, and OpenAI, plus an MCP integration for Cursor.

Developer Experience Score (9.0/10): Two official SDKs with TypeScript support, interactive playground, comprehensive docs, and MCP integration provide excellent developer experience, just slightly behind SerpAPI's eight-language SDK coverage.

Search Capabilities

Exa maintains a proprietary search index built for AI applications rather than scraping Google or Bing. The embeddings-based index covers web content, research papers, news, GitHub repositories, tweets, and PDFs with category filters and date filtering. The API provides multiple search modes (neural for semantic matching, keyword for traditional results, auto for intelligent combination) plus endpoints for content retrieval, similarity search, and automated research.

Search Capabilities Score (8.0/10): Proprietary AI-powered semantic search offers unique capabilities for meaning-based queries, though it can't replicate specific search engine results like SerpAPI.

Free Tier Generosity

The free tier provides $10 in credits with no expiration and no credit card requirement. At $5 per 1,000 searches for standard queries (1-25 results), this translates to 2,000 free searches. You can allocate credits across other endpoints (research/answers at 2,000 calls, content retrieval at 10,000 pages). All features work on the free tier including semantic search, similarity search, and AI-powered summaries.

Free Tier Score (9.0/10): 2,000 one-time free searches with no expiration places Exa among the most generous free tiers, second only to Scrapingdog's 11,000 credits.

Choose Exa when you need semantic search

Exa works best for AI agents, RAG systems, or LLM-powered research tools that benefit from meaning-based search. Voice AI companions leverage sub-2-second response times for fast web retrieval. Research platforms use the structured research endpoint to generate reports with citations. Chatbots get clean, parsed HTML without dealing with CAPTCHA.

Skip Exa if you need exact Google SERP replication for SEO monitoring or Google Shopping data for price comparison. Watch the 5x price jump when requesting 26-100 results per query ($0.025 per search approaches premium pricing without multi-engine coverage).

Tavily

Tavily is a search API optimized for LLMs and RAG applications. Instead of returning URLs with snippets, Tavily aggregates content from up to 20 sources per query, ranks them using proprietary AI, and delivers parsed content ready for LLM consumption.

Speed

Our 50-query benchmark measured 1.885s average response time with 100% success rate. Response times ranged from 1.583s to 2.897s. The response time reflects that Tavily fetches, parses, and ranks content before responding, while traditional SERP APIs return URLs and leave content extraction to you. Tavily's published benchmarks focus on accuracy (93.3% on SimpleQA) rather than raw speed.

Speed Score (8.5/10): At 1.885s average, Tavily ranks third among tested APIs, delivering strong performance considering it aggregates and parses content from multiple sources.

Pricing

Tavily's pricing uses a credit-based system. Pay-as-you-go charges $0.008 per credit. Basic search consumes 1 credit per request ($8 per 1,000 searches), while advanced search uses 2 credits. Subscription tiers start at $30/month for 4,000 credits ($0.0075 per credit) and scale down to $0.005 per credit on the Growth plan.

Pricing Score (7.0/10): At $8 per 1,000 searches on pay-as-you-go, Tavily offers mid-tier pricing between premium options like SerpAPI ($15) and budget options like Scrapingdog ($0.20).

Developer Experience

Tavily provides SDKs for Python (tavily-python) and JavaScript/TypeScript (@tavily/core). Setup takes about 5 minutes:

from tavily import TavilyClient

client = TavilyClient(api_key="tvly-YOUR_API_KEY")
response = client.search("coffee")

for result in response["results"]:
    print(result["title"])

Tavily's Playground generates real-time code as you build queries. Adjust parameters like search depth and result count, and the right panel updates with Python, JavaScript, or cURL code. Documentation includes integration guides for LangChain, LlamaIndex, OpenAI, Anthropic, Vercel AI, and Zapier.

Developer Experience Score (8.5/10): Two official SDKs, real-time code generation in the playground, comprehensive integration guides, and MCP support provide strong developer experience, comparable to Exa.

Search Capabilities

Tavily uses proprietary AI ranking to aggregate content from up to 20 sources per query. The API provides four endpoints: search (factual queries with AI ranking), extract (clean content from URLs), map (page discovery), and crawl (combined mapping and extraction). Search supports topic filters (general, news, finance), time range filters (day, week, month, year), and domain inclusion/exclusion.

Search Capabilities Score (7.5/10): AI-powered content aggregation with 93.3% accuracy on SimpleQA benchmarks provides strong factual search, though it can't replicate specific search engine results like SerpAPI or offer semantic search like Exa.

Free Tier Generosity

The free tier provides 1,000 credits per month with no credit card requirement. At 1 credit per basic search, this translates to 1,000 free searches monthly. You can allocate credits across advanced searches (2 credits), content extraction (1 credit per 5 URLs), or mapping (1 credit per 10 pages). All features work on the free tier with 100 requests per minute rate limiting.

Free Tier Score (8.0/10): 1,000 recurring monthly credits places Tavily among the more generous free tiers, better than SerpAPI (250/month) or Scrapingbee (66 total) but less than Exa (2,000 one-time) or Scrapingdog (11,000).

Choose Tavily for AI-optimized search

Tavily fits RAG applications, AI agents, or LLM-powered assistants that need factual, grounded web information. Chatbots get ranked content with source citations, research automation tools use the crawl endpoint to map entire domains, and content aggregation platforms pull clean text without handling CAPTCHA. The 93.3% accuracy on factual benchmarks and automatic content filtering reduce manual validation needs.

Skip Tavily for exact Google or Bing SERP layouts needed in SEO monitoring or structured shopping results for price comparison. Applications needing only URLs without content extraction should consider lighter-weight SERP APIs.

Scrapingdog

Scrapingdog is a web scraping service offering SERP API access alongside scrapers for Amazon, LinkedIn Jobs, Instagram, and other platforms. The service provides a Google Search API and a Universal Search API that aggregates and de-duplicates results across Google, Bing, Yahoo, and DuckDuckGo.

Speed

Our 50-query benchmark measured 1.700s average response time with 100% success rate. Response times ranged from 0.709s to 4.432s. The measured performance contradicts Scrapingdog's published benchmarks, which claim 1.25 second response times for Google SERP requests, though our 1.700s average is reasonably close.

Speed Score (9.0/10): At 1.700s average, Scrapingdog ranks second among tested APIs, delivering strong performance at the lowest price point.

Pricing

Scrapingdog's pricing uses a credit-based system with monthly subscriptions. Plans start at $40/month for 200,000 credits (Light) and scale to $30,000/month for 1 billion credits (Nova Pro). At 1 credit per Google search, the cost per 1,000 searches ranges from $0.20 (Light) to under $0.06 at high-volume tiers.

Pricing Score (10.0/10): At $0.20-$0.06 per 1,000 searches, Scrapingdog offers the cheapest pricing among all tested APIs, making it ideal for high-volume projects.

Developer Experience

Scrapingdog's documentation covers authentication, error codes, and endpoint references, though some pages returned 404 errors during testing. The service lacks official SDKs with helper functions found in competing APIs, requiring direct REST API calls using standard HTTP libraries. The dashboard's request builder generates code in cURL, Python, Java, Node.js, Ruby, PHP, Go, and C# as you configure parameters.

Setup takes approximately 5-10 minutes:

import requests

url = "https://api.scrapingdog.com/google"
params = {
    "api_key": "YOUR_API_KEY",
    "query": "coffee",
    "results": 10
}

response = requests.get(url, params=params)
data = response.json()

for result in data.get("organic_results", []):
    print(result["title"])

Developer Experience Score (5.0/10): No official SDKs, documentation gaps with 404 errors, and REST-only integration give Scrapingdog the weakest developer experience among tested APIs.

Search Capabilities

Scrapingdog's Google Search API provides access to Google web search with organic results, People Also Ask sections, related searches, and multiple search types (web, images, news, shopping). The Universal Search API aggregates and de-duplicates results from Google, Bing, Yahoo, and DuckDuckGo in a single request. Beyond search engines, Scrapingdog provides dedicated scrapers for Amazon, Walmart, eBay, LinkedIn Jobs, Instagram, and other platforms.

Search Capabilities Score (6.0/10): Coverage of 4 search engines via Universal Search API with de-duplication provides decent breadth, though significantly less than SerpAPI's 20+ platforms.

Free Tier Generosity

The free tier provides 1,000 request credits with no expiration and no credit card requirement. Optional social media promotions add 4,000 credits for following on X and 6,000 for posting a LinkedIn review, bringing the total to 11,000 credits. Free tier users access all API endpoints including the Universal Search API, JavaScript rendering, geo-targeting, and AI extraction features.

Free Tier Score (10.0/10): 11,000 total free credits (with promotions) is the most generous free tier among all tested APIs, ideal for extensive evaluation and small projects.

Choose Scrapingdog for cost-optimized scraping

Scrapingdog fits high-volume projects where per-request cost matters more than response time. Data collection pipelines requiring millions of monthly searches benefit from pricing that drops to $0.00006 per request at scale. Lead generation tools can consolidate Google search results alongside LinkedIn profiles, Instagram accounts, or e-commerce platforms through a single provider. The Universal Search API's de-duplication feature benefits SEO tools comparing rankings across multiple search engines without manual result matching.

Skip Scrapingdog for projects requiring official SDKs and comprehensive documentation. The REST-only API and documentation gaps observed during testing make integration more challenging than competitors with polished SDKs.

Scrapingbee

Scrapingbee is a general-purpose web scraping service that includes Google Search API functionality. Acquired by Oxylabs in June 2025, it targets developers needing headless browser rendering, JavaScript execution, and proxy rotation across multiple websites including e-commerce platforms, social media sites, and search engines.

Speed

Our 50-query benchmark measured 7.257s average response time with 100% success rate. Response times ranged from 1.453s to 22.978s, creating a 16x performance spread that makes response time prediction difficult. Competitor benchmarks from Scrapingdog show 90% success rate with 16.28 second average response time for Google SERP requests, confirming high variability.

Speed Score (2.0/10): At 7.257s average, Scrapingbee delivered the slowest response times among all tested APIs, with extreme variability reflecting its positioning as a general web scraper rather than SERP-optimized service.

Pricing

Scrapingbee's pricing uses a credit-based system where Google Search API requests cost 15 credits per call. Monthly plans start at $49.99 for 250,000 credits. For 1,000 Google searches at 15 credits per request, costs range from $0.56 to $1.44 depending on tier. Variable credit costs create billing unpredictability as enabling premium proxies increases request cost from 5 to 25 credits (5x multiplier).

Pricing Score (8.5/10): At $0.56-$1.44 per 1,000 searches, Scrapingbee offers competitive pricing in the mid-to-low range, more affordable than SerpAPI ($15) or Exa ($5) but slightly more expensive than Scrapingdog ($0.20).

Developer Experience

Scrapingbee's documentation provides endpoint references, authentication guides, and code examples across Python, Node.js, Java, Ruby, PHP, Go, and C#. The request builder interface shows real-time credit costs and generates code in multiple languages as you configure parameters. Scrapingbee provides official Python and Node.js SDKs that wrap authentication and parameter formatting.

Setup takes approximately 5-10 minutes:

import requests

url = "https://app.scrapingbee.com/api/v1"
params = {
    "api_key": "YOUR_API_KEY",
    "search": "coffee",
    "search_type": "search"
}

response = requests.get(url, params=params)
data = response.json()

for result in data.get("organic_results", []):
    print(result["title"])

Developer Experience Score (7.0/10): Two official SDKs, comprehensive documentation, and a request builder with real-time cost estimation provide decent developer experience, though credits charged for failed requests create cost uncertainty.

Search Capabilities

Scrapingbee's Google Search API provides access to Google web search only with structured JSON including organic results, local results, ads, related queries, images, news results, knowledge graph data, and top stories. Scrapingbee supports multiple Google search types including web, news, Maps, and image search. The platform provides dedicated scrapers for Amazon, Walmart, and ChatGPT API alongside general HTML scraping.

Search Capabilities Score (4.0/10): Google-only coverage limits Scrapingbee to a single search engine, significantly less than SerpAPI (20+ engines), Scrapingdog (4 engines), or the proprietary search capabilities of Exa and Tavily.

Free Tier Generosity

The free tier provides 1,000 API credits with no credit card required and no time expiration. At 15 credits per Google search, this translates to approximately 66 free searches. The 1,000 credits cover diverse use cases including HTML scraping (5 credits), Amazon scraping (15 credits), and Walmart scraping (15 credits), with all platform features available including JavaScript rendering, geo-targeting, and premium proxies at higher credit costs.

Free Tier Score (2.0/10): 66 total free searches is the least generous free tier among all tested APIs, limiting evaluation and making it difficult to thoroughly test the service before committing.

Choose Scrapingbee for unified web scraping

Scrapingbee fits projects requiring both SERP data and general web scraping in a unified platform. Teams already scraping e-commerce sites, social media profiles, or JavaScript-heavy web applications can consolidate Google search results through the same service. Projects with flexible latency requirements can tolerate 7.257-second average response times and high variability for data collection pipelines, batch processing systems, or research workflows prioritizing breadth over speed.

Skip Scrapingbee for dedicated SERP API needs. The 66-search free tier limits evaluation, Google-only coverage restricts multi-engine analysis, and 7.257-second average response times with 23-second worst-case performance make it unsuitable for real-time voice assistants or live autocomplete suggestions.

SerpAPI vs Tavily

SerpAPI and Tavily represent different approaches to SERP data. SerpAPI scrapes existing search engines like Google and Bing, returning their results as structured JSON. Tavily runs its own search engine that aggregates content from up to 20 sites and ranks it using proprietary AI. This architectural difference determines which one fits your use case.

Mirror Google or aggregate multiple sources?

When you query SerpAPI for Google results, you get exactly what appears on Google with all its ranking signals, personalization, and algorithmic biases. When you query Amazon, you get Amazon's product listings ranked by Amazon's algorithm. This matters for SEO monitoring, competitor analysis, and applications that need to replicate what users see on specific platforms. SerpAPI supports 20+ search engines with dedicated endpoints for each.

Tavily doesn't scrape existing engines. It aggregates and ranks content from multiple sources per query, scoring relevance with AI and formatting results for LLM consumption. You can't get "Bing results" from Tavily because Tavily is the search engine. This matters for AI agents and RAG applications where answer accuracy matters more than which search engine provided it.

Our benchmarks measured SerpAPI at 0.072 seconds average response time versus Tavily's 2.327 seconds. The 32x speed difference reflects Tavily's extra processing: searching up to 20 sites, applying AI ranking, extracting content, and formatting for LLM context windows. SerpAPI returns Google's results as-is with minimal overhead.

Choose SerpAPI when you need specific search engines

Build with SerpAPI when you need to replicate what users see on specific platforms. SEO monitoring tools that track keyword rankings across Google, Bing, and other engines require SerpAPI's multi-engine support. Price comparison tools need Amazon, eBay, and Walmart product listings. Academic research applications can use SerpAPI's Scholar endpoint instead of building separate scrapers.

The 0.072-second response time fits latency-sensitive applications like browser extensions, real-time dashboards, and user-facing search features. When your application needs to feel instant, the 32x speed advantage matters.

Choose Tavily when you're feeding LLMs

Build with Tavily when creating AI agents, RAG applications, or LLM-powered tools that need accurate information rather than specific search engine results. Tavily's 93.3% accuracy on SimpleQA and LLM-optimized formatting reduce the post-processing needed to integrate search into your agent's workflow.

The multi-source aggregation helps when you care about answer quality over result provenance. Instead of scraping Google's top result and hoping it's accurate, Tavily retrieves up to 20 sources and ranks them by relevance. If your application shows users "here's what I found about X" rather than "here are Google's top 10 results for X," Tavily's AI-scored relevance provides better raw material.

The 2.327-second response time becomes acceptable when the alternative is making multiple follow-up requests to scrape and parse content from traditional SERP results.

How to Choose the Right SERP API

SerpAPI works best when you need exact Google (or Bing, or Amazon) results with sub-100ms response times. The $15 per 1,000 searches pays for 20+ search engine coverage and infrastructure that handles CAPTCHA solving automatically. SEO monitoring tools and price comparison platforms that need traditional SERP layouts benefit from the 0.072-second average we measured.

Exa fits AI applications needing semantic search over a proprietary index. The $5 per 1,000 searches includes embeddings and meaning-based matching that keyword APIs don't provide. At 1.189 seconds average, Exa delivers the fastest semantic search among tested options. The 2,000 free searches make evaluation straightforward.

Tavily targets RAG pipelines requiring factual, multi-source content. The service aggregates up to 20 sites per query and formats results for LLM consumption at $5-$8 per 1,000 searches. The 1,000 monthly free credits that reset support sustained prototyping. Choose Tavily when accuracy matters more than speed.

Scrapingdog offers the lowest per-request pricing at scale, dropping to $0.000058 per request for high-volume users. Our 3.762-second average contradicted their claimed 1.25-second benchmark, but the cost advantage persists for budget-focused projects. The Universal Search API de-duplicates across four engines in single requests.

Scrapingbee makes sense for teams already scraping e-commerce or general websites who want to consolidate SERP access through one provider. The 5.411-second average and high variability make it unsuitable for latency-sensitive applications. At 66 free searches versus competitors offering 250-2,000, the free tier limits evaluation depth.

Frequently Asked Questions

What is a SERP API?

A SERP API provides programmatic access to search engine results through structured JSON responses. Instead of manually searching Google and scraping HTML, you send an HTTP request with query parameters and receive formatted data including organic results, ads, knowledge panels, and related searches. These APIs handle CAPTCHA solving and proxy rotation automatically.

Are SERP APIs legal?

SERP APIs operate in a legal gray area. In the United States, hiQ Labs v. LinkedIn (2019) established that scraping publicly accessible data doesn't violate the Computer Fraud and Abuse Act. Search engine terms of service typically prohibit automated access, but enforcement focuses more on technical countermeasures than legal action. SerpAPI offers U.S. Legal Shield on paid plans, covering legal fees if you face action for scraping public search data.

How do SERP APIs avoid getting blocked?

SERP APIs use rotating residential proxies, CAPTCHA solving services, and request pattern randomization to avoid detection. They distribute requests across thousands of IP addresses that appear as regular user traffic. When search engines implement new anti-bot measures, the APIs update their infrastructure to maintain access. This is why using an API makes more sense than building your own scraping system.

Can I use SERP APIs for commercial projects?

Yes, all five APIs support commercial use. Check your plan's terms of service for specific restrictions. Most APIs prohibit reselling raw SERP data to competitors but allow using the data in your own products. SerpAPI's pricing page and Tavily's documentation specify commercial terms. Free tiers typically allow commercial use with volume limitations.

Testing Methodology

Back to comparison results

We tested each API with fifty standardized queries: "python tutorials", "best coffee makers 2025", "machine learning", "weather forecast" etc. The Python script measured response time from request to complete JSON parsing and verified successful responses. Each test ran once per API to simulate real-world usage.

import requests
import time
import os
import json
from dotenv import load_dotenv
from urllib.parse import quote_plus

load_dotenv()

def test_serpapi(query):
    start = time.time()
    response = requests.get("https://serpapi.com/search", params={
        "q": query,
        "api_key": os.getenv("SERPAPI_KEY")
    })
    elapsed = time.time() - start
    return elapsed, response.status_code == 200

def test_exa(query):
    start = time.time()
    response = requests.post("https://api.exa.ai/search",
        headers={"x-api-key": os.getenv("EXA_KEY")},
        json={"query": query, "numResults": 10})
    elapsed = time.time() - start
    return elapsed, response.status_code == 200

def test_tavily(query):
    start = time.time()
    response = requests.post("https://api.tavily.com/search",
        json={"api_key": os.getenv("TAVILY_KEY"), "query": query})
    elapsed = time.time() - start
    return elapsed, response.status_code == 200

def test_scrapingdog(query):
    start = time.time()
    response = requests.get("https://api.scrapingdog.com/google", params={
        "api_key": os.getenv("SCRAPINGDOG_KEY"),
        "query": query
    })
    elapsed = time.time() - start
    return elapsed, response.status_code == 200

def test_scrapingbee(query):
    start = time.time()
    encoded_query = quote_plus(query)
    response = requests.get("https://app.scrapingbee.com/api/v1/", params={
        "api_key": os.getenv("SCRAPINGBEE_KEY"),
        "url": f"https://www.google.com/search?q={encoded_query}",
        "render_js": "false",
        "custom_google": "True"
    })
    elapsed = time.time() - start
    return elapsed, response.status_code == 200

queries = [
    "python tutorials", "best coffee makers 2025", "machine learning",
    "weather forecast", "electric cars", "react hooks tutorial",
    "best restaurants near me", "data science courses", "yoga poses",
    "home workout routines", "javascript async await", "docker compose",
    "kubernetes tutorial", "aws lambda pricing", "terraform vs ansible",
    "best noise cancelling headphones", "meal prep ideas", "sourdough recipe",
    "running shoes 2025", "meditation techniques", "typescript generics",
    "postgres indexing", "redis caching", "mongodb aggregation",
    "git rebase vs merge", "vim shortcuts", "linux commands",
    "css flexbox", "tailwind css", "nextjs routing",
    "graphql vs rest", "microservices architecture", "api gateway",
    "oauth2 flow", "jwt authentication", "password hashing",
    "sql injection prevention", "xss attack", "csrf protection",
    "seo best practices", "meta tags", "sitemap xml",
    "google analytics", "conversion rate optimization", "a/b testing",
    "email marketing", "social media strategy", "content marketing",
    "project management tools", "agile methodology", "scrum vs kanban"
]

apis = {
    "SerpAPI": test_serpapi,
    "Exa": test_exa,
    "Tavily": test_tavily,
    "Scrapingdog": test_scrapingdog,
    "Scrapingbee": test_scrapingbee
}

results = {}

for api_name, test_func in apis.items():
    print(f"\nTesting {api_name}...")
    times = []
    successes = 0
    for i, query in enumerate(queries, 1):
        try:
            elapsed, success = test_func(query)
            times.append(elapsed)
            if success:
                successes += 1
            print(f"  {i}/50: {query[:30]:30s} - {elapsed:.3f}s - {'✓' if success else '✗'}")
        except Exception as e:
            print(f"  {i}/50: {query[:30]:30s} - ERROR: {str(e)[:50]}")
            times.append(0)

    avg_time = sum(times) / len(times) if times else 0
    success_rate = (successes / len(queries)) * 100
    fastest = min(times) if times else 0
    slowest = max(times) if times else 0

    results[api_name] = {
        "avg_time": round(avg_time, 3),
        "success_rate": round(success_rate, 1),
        "fastest": round(fastest, 3),
        "slowest": round(slowest, 3),
        "total_queries": len(queries),
        "successful_queries": successes
    }

    print(f"\n{api_name} Results:")
    print(f"  Average: {avg_time:.3f}s")
    print(f"  Success Rate: {success_rate}%")
    print(f"  Fastest: {fastest:.3f}s")
    print(f"  Slowest: {slowest:.3f}s")

# Save results to JSON
with open("test_results.json", "w") as f:
    json.dump(results, f, indent=2)

# Print summary
print("\n" + "="*60)
print("FINAL SUMMARY")
print("="*60)
for api_name, data in results.items():
    print(f"{api_name:15s} {data['avg_time']:6.3f}s avg, {data['success_rate']:5.1f}% success")

We ran tests with a 100Mbps connection. Network latency impacts absolute times but relative performance remains consistent. Your results may vary based on location, query complexity, and API server load.

How to create a timeline: Excel, Word, PowerPoint, Google Sheets, Preceden, and JavaScript

RitzaCo — Thu, 13 Nov 2025 07:54:12 +0000

A timeline visualizes events chronologically, making it easier to understand project progress, historical sequences, or future plans. Whether you're tracking project milestones, creating a presentation, or documenting company history, you need the right tool to build your timeline.

This guide shows you how to create timelines using tools you likely already have:

Excel: Using SmartArt for basic document timelines.
Word: Using SmartArt for report and document timelines.
PowerPoint: Using SmartArt for presentation timelines.
Google Sheets/Docs: Using Timeline view for collaborative projects (requires Google Workspace).
Preceden: Using dedicated timeline software for professional results.
JavaScript: Using Bryntum to build custom timeline charts.

Each section is self-contained, so you can jump directly to the tool you want to use. Before diving into the how-to sections, you'll learn the key differences between timelines and Gantt charts to choose the proper visualization for your needs.

What Are the Differences Between a Timeline and a Gantt?

Timelines and Gantt charts both visualize events over time, but they serve different purposes. Understanding the difference helps you choose the right tool for your project.

What's a Timeline?

A timeline is a linear visualization that shows events in chronological order.

Timelines focus on dates and milestones, making them ideal for:

Historical events such as company founding, product launches, and industry evolution.
Project milestones like major deadlines and deliverables.
Event planning, like conference agendas.

A timeline answers the question: What happened when?

What's a Gantt Chart?

A Gantt chart shows when events happen, how long they take, and how they relate to each other.

Gantt charts display:

Task start and end dates.
Task duration.
Overlapping tasks.
Task dependencies – which tasks must finish before others can start.

A Gantt chart answers the questions: When does this happen, how long does it take, and what depends on it?

Timeline vs Gantt Chart: Which One Should You Use?

Choose a timeline when you want to communicate a sequence of events to an audience. Timelines work best for presentations, reports, and documentation where clarity matters more than detailed project management.

Choose a Gantt chart when you need to plan and track complex projects with multiple tasks, dependencies, and resources. Gantt charts work best for project managers coordinating teams and deliverables.

How to Create a Timeline in Excel?

Excel doesn't have a dedicated timeline chart. You can create a basic timeline using SmartArt graphics, which work in Excel desktop applications but not in Excel Online.

Step 1: Insert SmartArt

On the Excel application, follow these steps:

Go to the Insert tab.
Select the SmartArt option.
Then select "Process". You will see many options to describe a process, a timeline.
Choose the one that works for you, but for this tutorial, we will use the first one, "Basic process".

Step 2: Add Your Events

Excel displays a timeline with placeholder text. Click each text box and type your event name and date.

To add more events, you can use the SmartArt text utility, by typing on the "Enter" button for a new line, or the "+" button. A new item will appear on the right of your timeline.

Step 3: Format Your Timeline

Customize your timeline using the SmartArt Design tab. You can change the colors, apply SmartArt styles or change the layout.

For detailed formatting, select individual text boxes and use the Format tab to change fill colors, borders, or text properties.

How to Create a Timeline in PowerPoint

PowerPoint provides SmartArt graphics that make timeline creation straightforward. This feature is available in both PowerPoint desktop and PowerPoint Online.

Step 1: Insert SmartArt

In your PowerPoint presentation, follow these steps:
1. Click the Insert tab.
2. Click SmartArt.
3. Select "Process" from the left panel. You'll see various process diagrams, including timeline options.
4. Select the timeline style you prefer. For this guide, we'll use "Basic Process".

Step 2: Add Your Events

A timeline graphic will appear with text placeholders. Click on each text box and type your event information.

To add more events, you can use the SmartArt text utility by typing on the "Enter" button for a new line or the "+" button.

Step 3: Format Your Timeline

Customize your timeline using the SmartArt Design tab. You can change the colors, apply SmartArt styles, or change the layout.

For detailed formatting, select individual text boxes and use the Format tab to change fill colors, borders, or text properties.

How to Create a Timeline in Word?

Word provides SmartArt graphics for creating basic timelines in your documents. This method works in Word desktop applications (not Word Online).

Step 1: Insert SmartArt

On the Word application, follow these steps:

Go to the Insert tab.
Select the SmartArt option.
Then select "Process". You will see many options to describe a process, including timeline layouts.
Choose the one that works for you, but for this tutorial, we will use "Basic process".

Step 2: Add Your Events

A timeline graphic will appear with text placeholders. Click on each text box and type your event information - include both the event name and date.

To add more events, click the SmartArt border, then click the SmartArt Design tab > Add Shape. New boxes appear on your timeline.

To remove events, click a box and press Delete.

Step 3: Format Your Timeline

Customize your timeline using the SmartArt Design tab. You can change the colors, apply SmartArt styles, or change the layout.

For detailed formatting, select individual text boxes and use the Format tab to change fill colors, borders, or text properties.

How to Create a Timeline in Google Docs

Google Docs doesn't have a built-in timeline feature. You can create a timeline in Google Sheets using Timeline view, then add it to your document as an image.

Note: Timeline view is only available for Google Workspace accounts (Business Starter, Business Standard, Business Plus, Enterprise, Education editions). Free Google accounts do not have access to this feature.

Step 1: Prepare Your Data in Google Sheets

Create a new Google Sheet and set up your timeline data in columns:

Card Title,Start Date,End Date,Card Detail
Research & Planning,1/15/2025,2/1/2025,Initial research phase
Design Phase,2/15/2025,3/15/2025,UI/UX design
Development,4/15/2025,6/15/2025,Build features
Testing,6/15/2025,7/1/2025,QA and bug fixes
Launch,7/15/2025,7/15/2025,Product launch

Enter your dates in a standard format like 1/15/2025 or 01/15/2025. Google Sheets recognizes these as dates automatically and right-aligns them in the cells.

Step 2: Insert Timeline View

In your Google Sheet, follow these steps:

Go to the Insert menu.
Click Timeline.

Select your data range, including all columns with headers.

Click Ok after validating the selection.

Google Sheets creates an interactive timeline view showing your tasks as cards on a horizontal timeline.

On the right panel, you can customize the timeline by modifying the card title, the card color, or the display.

Step 4: Embed in Google Docs

Google Docs doesn't support embedding timelines directly from Google Sheets. To add your timeline to a document:

In Google Sheets, click File -> Download -> PDF Document.
Convert the PDF to an image using a tool like pdf2png.com or CloudConvert.
In Google Docs, click Insert -> Image -> Upload from computer.
Select your converted timeline image.

How to Create a Timeline in Preceden?

Preceden is a dedicated timeline software explicitly designed for creating professional timelines. The free plan lets you create timelines manually, while paid plans allow you to import spreadsheet data for faster timeline creation.

Step 1: Create a new timeline

Visit preceden.com and sign up for an account. On the dashboard, click the "+Create Manually" button.

Enter a name for your timeline and click "Create Timeline →".

Step 2: Add events to your timeline

Preceden displays your blank timeline. Click the "+ Add Event" button.

Enter your event details, such as:

Event name.
Start date.
End date (optional for milestone events).

Click Save. Your event appears on the timeline immediately.

Repeat this process to add more events. Alternatively, click "Bulk Add" to enter multiple events at once using a spreadsheet-style interface.

Step 3: Use the timeline

Preceden allows you to export the timeline in many formats, such as PDF or Images, or embed it if you need to display the chart on a website.

How to Create a Timeline in JavaScript

Creating timelines in web applications requires a JavaScript library that handles rendering, interactivity, and data management. While you could build a timeline from scratch using HTML canvas or SVG, production-quality timelines need features like drag-and-drop, zoom controls, and responsive design.

For this guide, you will learn how to create a timeline using the Bryntum Scheduler Pro component, which has easy setup and provides complete timeline functionality. Bryntum offers trial packages so you can test the library before purchasing. If you're interested in other JavaScript timeline options, you can also check vis-timeline (open source) and Timeline.js (free for storytelling timelines).

Prerequisites

Before creating your timeline, you need:

Node.js and npm are installed on your computer.
A Bryntum account (free trial available).

Step 1: Initialize your project

Create a new project directory and initialize an npm project:

mkdir timeline-project
cd timeline-project
npm init -y

Step 2: Set up npm authentication

Bryntum packages require authentication. Follow the npm repository login guide to authenticate your npm client.

Then configure npm to access Bryntum packages by following the npm configuration guide.

Step 3: Install Bryntum Scheduler

Install Bryntum Scheduler. You can use the trial version or purchase a license. Follow the installation guide for your chosen version.

For the trial version:

npm install @bryntum/scheduler@npm:@bryntum/scheduler-trial@6.3.3

Step 4: Create your HTML file

Inside the project directory, create an index.html file with the timeline container:

<!DOCTYPE html>
<html lang="en">
<head>
    <meta charset="UTF-8">
    <meta name="viewport" content="width=device-width, initial-scale=1.0">
    <title>Timeline Chart</title>
    <link rel="stylesheet" href="node_modules/@bryntum/scheduler/scheduler.stockholm.css">
    <style>
        body {
            font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', sans-serif;
            margin: 20px;
            background: #f5f5f5;
        }
        #timeline-container {
            height: 500px;
            background: white;
            border-radius: 8px;
            box-shadow: 0 2px 8px rgba(0,0,0,0.1);
        }
    </style>
</head>
<body>
    <div id="timeline-container"></div>
    <script type="module" src="app.js"></script>
</body>
</html>

Step 5: Create your JavaScript timeline

Create an app,js file with the following timeline configuration:

import { Scheduler, DateHelper } from './node_modules/@bryntum/scheduler/scheduler.module.js';

// Define your timeline events
const events = [
    {
        id: 1,
        name: 'Research & Planning',
        startDate: new Date(2025, 0, 15),
        endDate: new Date(2025, 1, 1),
        resourceId: 1
    },
    {
        id: 2,
        name: 'Design Phase',
        startDate: new Date(2025, 1, 15),
        endDate: new Date(2025, 2, 15),
        resourceId: 1
    },
    {
        id: 3,
        name: 'Development',
        startDate: new Date(2025, 3, 15),
        endDate: new Date(2025, 5, 15),
        resourceId: 1
    },
    {
        id: 4,
        name: 'Testing',
        startDate: new Date(2025, 5, 15),
        endDate: new Date(2025, 6, 1),
        resourceId: 1
    },
    {
        id: 5,
        name: 'Launch',
        startDate: new Date(2025, 6, 15),
        endDate: new Date(2025, 6, 15),
        resourceId: 1
    }
];

// Create a single resource (timeline row)
const resources = [
    { id: 1, name: 'Project Timeline' }
];

// Initialize the timeline
const scheduler = new Scheduler({
    appendTo: 'timeline-container',
    resources: resources,
    events: events,
    startDate: new Date(2025, 0, 1),
    endDate: new Date(2025, 7, 1),
    viewPreset: 'weekAndMonth',
    columns: [
        { text: 'Timeline', field: 'name', width: 150 }
    ],
    features: {
        eventDrag: true,
        eventResize: true,
        eventEdit: true,
        eventTooltip: true
    },
    rowHeight: 60,
    barMargin: 10
});

Step 6: Run your timeline

Open index.html in a web browser.

You'll see an interactive timeline where you can:

Drag events to change their dates.
Resize events to adjust duration.
Click events to edit details.
Hover over events to see tooltips.

Conclusion

This guide showed you how to create timelines in Excel, Word, PowerPoint, Google Docs, Preceden, and JavaScript using Bryntum Scheduler Pro.

Google Sheets and Preceden offer the best experience because they use your spreadsheet data directly and let you easily customize colors. Microsoft tools use SmartArt, which is simple to use but requires manual updates for each event.

For occasional timelines in documents or presentations, SmartArt works fine. For professional timelines with many events or frequent updates, use Preceden. And if you're a developer building a dashboard or tool that needs embedded timelines, use Bryntum Scheduler Pro for complete control over appearance and functionality.

CRM app with Node.js, Replit, and MongoDB

RitzaCo — Tue, 20 Sep 2022 09:09:47 +0000

In this tutorial, we'll use Node.js on Replit and a MongoDB database to build a basic CRUD (Create, Read, Update, Delete) CRM (Customer Relationship Management) application. A CRM lets you store information about customers to help you track the status of every customer relationship. The application will be able to store and edit customer details, as well as keep notes about them.

This tutorial won't be covering the basics of Node.js, but each line of code will be explained in detail.

Setting up

All the code for our application will be written and hosted on Replit, so you won't need to install any additional software on your computer.

If you don't already have a Replit account, create one now.

Next, head over to MongoDB Atlas and hit the "Try free" button. You should then sign up, clicking the "Get started free" button to complete the process.

MongoDB Atlas is a fully managed Database-as-a-Service. It provides a document database (often referred to as NoSQL), as opposed to a more traditional relational database like PostgreSQL.

Creating a cluster

Once you've signed up with MongoDB Atlas, under "Shared Clusters", press the "Create a Cluster" button.

Select a provider and a region. For the purposes of this tutorial, we'll choose "Google Cloud Platform" as the provider and "Iowa (us-central1)" as the region, although it should work regardless of the provider and region.

Give your cluster a name. Note that once you've created your cluster, you won't be able to change the name. Click "Create Cluster".

After a bit of time, your cluster will be created. Once it's available, click on “Database Access” under the Security heading in the left-hand column and then click "Add New Database User". You need a database user to store and retrieve data. Enter a username and password for the user and make a note of those details - you'll need them later. Select “Read and write to any database” as the user privilege. Hit "Add User" to complete this step.

Next, you need to allow network access to the database. Click on "Network Access" in the left-hand column, and “Add IP Address”. Because we won't have a static IP from Replit, we're just going to allow access from anywhere - don't worry, the database is still secured with the username and password you created earlier. In the pop-up, click "Allow Access From Anywhere" and then "Confirm".

Now select "Database" under "Deployment" in the left-hand column. Click on "Connect" and select “Connect Your Application”. This will change the pop-up view. Copy the "Connection String" as you will need it shortly to connect to your database from Replit. It will look something like this: mongodb+srv://<username>:<password>@cluster0-zrtwi.gcp.mongodb.net/test?retryWrites=true&w=majority

Creating a repl and connecting to the database

Now we need a Node.js repl to write the code necessary to connect to our shiny new database. Navigate to Replit and create a new repl, selecting "Node.js" as the language.

A great thing about Replit is that it makes projects public by default. This makes it easy to share and it's great for collaboration and learning, but we have to be careful not to make our database credentials available on the open internet.

To solve this problem, we'll use Replit's environment variables, as we have done in previous tutorials. Replit allows us to set secret environment variables through the "Secrets (Environment variables)" menu option.

Open the "Secrets" menu option. There you will be able to set environment variables for your repl. Set the key as the name of your environment variable to MONGO_USERNAME. Set the value as your database username, then click "Add new secret". Then create a new key called MONGO_PASSWORD with its value set to your database password, and click "Add new secret". It should look something like:

Replace username and password with your database username and password.

Now that we have credentials set up for the database, we can move on to connecting to it in our code.

MongoDB is kind enough to provide a client that we can use. To test out our database connection, we're going to insert some customer data into our database. In your index.js file (created automatically and found under the Files pane), add the following code:

const MongoClient = require('mongodb').MongoClient;
const mongo_username = process.env['MONGO_USERNAME']
const mongo_password = process.env['MONGO_PASSWORD']

const uri = `mongodb+srv://${mongo_username}:${mongo_password}@cluster0-zrtwi.gcp.mongodb.net/crmdb?retryWrites=true&w=majority`;
const client = new MongoClient(uri, { useNewUrlParser: true });

Let's break this down to see what is going on:

Line 1 adds the dependency for the MongoDB Client. Replit makes things easy by installing all the dependencies for us, so we don't have to use something like npm to do it manually.
In Line 2 & 3, we retrieve our MongoDB username and password from the environment variables that we set up earlier.
Line 5 has a few very important details that we need to get right:
- Replace the section between the @ and the next / with the same section of your connection string from MongoDB that we copied earlier. You may notice the ${mongo_username} and ${mongo_password} before and after the colon near the beginning of the string. These are called template literals. Template literals allow us to put variables in a string, which Node.js will then replace with the actual values of the variables.
- Note crmdb after the / and before the ?. This will be the name of the database that we will be using. MongoDB creates the database for us if it doesn't exist. You can change this to whatever you want to name the database, but remember what you changed it to for future sections of this tutorial.
Line 6 creates the client that we will use to connect to the database.

Making a user interface to insert customer data

We're going to make an HTML form that will capture the customer data and send it to our Replit code, which will then insert it into our database.

In order to actually present and handle an HTML form, we need a way to process HTTP GET and POST requests. The easiest way to do this is to use a web application framework. A web application framework is designed to support the development of web applications - it gives you a standard way to build your application and lets you get to building your application fast without having to do the boilerplate code.

A really simple, fast, and flexible Node.js web application framework is Express, which provides a robust set of features for the development of web applications.

The first thing we need to do is add the dependencies we need. Right at the top of your index.js file (above the MongoDB code), add the following lines:

let express = require('express');
let app = express();
let bodyParser = require('body-parser');
let http = require('http').Server(app);

app.use(bodyParser.json())
app.use(bodyParser.urlencoded({ extended: true }));

Let's break this down.

Line 1 adds the dependency for Express. Replit will take care of installing it for us.
Line 2 creates a new Express app that will handle incoming requests.
Line 3 adds a dependency for 'body-parser'. This is needed for the Express server to be able to handle the data that the form will send, and give it to us in a useful format to use in the code.
Line 4 adds a dependency for a basic HTTP server.
Line 6 & 7 tell the Express app which parsers to use on incoming data. This is needed to handle form data.

Next, we need to add a way for the Express app to handle an incoming request and give it to us in the form we need. Add the following lines of code below the code you added above:

app.get('/', function (req, res) {
  res.sendFile('/index.html', {root:'.'});
});

app.get('/create', function (req, res) {
  res.sendFile('/create.html', {root:'.'});
});

app.get tells Express that we want it to handle a GET request.
'/' tells Express that it should respond to GET requests sent to the root URL. A root URL looks something like https://crm.hawkiesza.repl.co - note that there are no slashes after the URL.
'/create' tells Express that it should respond to GET requests sent to the /create endpoint after the root URL, i.e. https://crm.hawkiesza.repl.co/create.
res.sendFile tells Express to send the given file as a response.

Before the server will start receiving requests and sending responses, we need to tell it to run. Add the following code below the previous line:

app.set('port', process.env.PORT || 5000);
http.listen(app.get('port'), function() {
    console.log('listening on port', app.get('port'));
});

Line 1 tells Express to set the port number to either a number defined as an environment variable, or 5000 if no definition was made.
Line 2-4 tells the server to start listening for requests.

Now we have an Express server listening for requests, but we haven't yet built the form that it needs to send back if it receives a request.

Make a new file called index.html and paste the following code into it:

<!DOCTYPE html>
<html lang="en">
<body>
<form action="/create" method="GET">
  <input type="submit" value="Create">
</form>

</body>
</html>

This is a straightforward bit of HTML that puts a single button on the page. When this button is clicked, it sends a GET request to /create, which the server will then respond to according to the code that we wrote above - in our case, it will send back the create.html file which we will define now.

Make a new file called create.html and paste the following into it:

<!DOCTYPE html>
<html lang="en">
<body>

<h2>Customer details</h2>

<form action="/create" method="POST">
  <label for="name" >Customer name *</label><br>
  <input type="text" id="name" name="name" class="textInput" placeholder="John Smith" required>
  <br>
  <label for="address" >Customer address *</label><br>
  <input type="text" name="address" class="textInput" placeholder="42 Wallaby Way, Sydney" required>
  <br>
  <label for="telephone" >Customer telephone *</label><br>
  <input type="text" name="telephone" class="textInput" placeholder="+275554202" required>
  <br>
  <label for="note" >Customer note</label><br>
  <input type="text" name="note" class="textInput" placeholder="Needs a new pair of shoes">
  <br><br>
  <input type="submit" value="Submit">
</form>

</body>
</html>

We won't go in-depth into the above HTML. It is a very basic form with four fields (name, address, telephone, note) and a submit button, which creates an interface that will look like the one below.

When the user presses the submit button, a POST request is made to /create with the data in the form - we still have to handle this request in our code as we're currently only handling a GET request to /.

If you click the "Run" button now, a new window should appear on the right that displays the "Create" button we defined just now in index.html. To see the form, you can also navigate to https://<repl_name>.<your_username>.repl.co, replacing with whatever you named your repl (but with no underscores or spaces) and with your repl username. You will be able to see this URL in the repl itself.

If you select "Create" and then fill in the form and hit submit, you'll get a response that says

Cannot POST /create

. This is because we haven't added the code that handles the form POST request, so let's do that.

Add the following code into your index.js file, below the app.get entry that we made above:

app.post('/create', function (req, res, next) {
  client.connect(err => {
    const customers = client.db("crmdb").collection("customers");

    let customer = {name: req.body.name, address: req.body.address, telephone: req.body.telephone, note: req.body.note};
    customers.insertOne(customer, function (err, res) {
      if (err) throw err;
      console.log("1 customer inserted");
    });
  })
  res.send('Customer created');
})

Line 1 defines a new route that listens for an HTTP POST request at /create.
Line 2 connects to the database. This happens asynchronously, so we define a callback function that will be called once the connection is made.
Line 3 creates a new collection of customers. Collections in MongoDB are similar to tables in SQL.
Line 5 defines customer data that will be inserted into the collection. This is taken from the incoming request. The form data is parsed using the parsers that we defined earlier and is then placed in the req.body variable for us to use in the code.
Line 6 inserts the customer data into the collection. This also happens asynchronously, and so we define another callback function that will get an error if an error occurred, or the response if everything happened successfully.
Line 7 throws an error if the above insert had a problem.
Line 8 gives us some feedback that the insert happened successfully.

If you run the repl now (you may need to refresh it) and submit the filled-in form, you'll get a message back that says "Customer created". If you look in your cluster in MongoDB and select the "Collections" button, you'll see a document has been created with the details that we submitted in the form.

Updating and deleting database entries

As a final step in this tutorial, we want to be able to update and delete database documents in our collection. To make things simpler, we're going to make a new HTML page where we can request a document and then update or delete it.

First, let's make the routes to our new page. In your index.js, add the following code below the rest of your routing code (i.e. before the MongoDB code):

app.get('/get', function (req, res) {
  res.sendFile('/get.html', {root: '.'});
});

app.get('/get-client', function (req, res) {
    client.connect(err => {
        client.db("crmdb").collection("customers").findOne({name: req.query.name}, function(err, result) {
          if (err) throw err;
          res.render('update', {oldname: result.name, oldaddress: result.address, oldtelephone: result.telephone, oldnote: result.note, name: result.name, address: result.address, telephone: result.telephone, note: result.note});
        });
      });
});

Line 1-3 tells Express to respond to incoming GET requests on /get by sending the get.html file which we will define below.
Line 5-12 tells Express to respond to incoming GET requests on /get-client.
- Line 7 makes a call to the database to fetch a customer by name. If there is more than one customer with the same name, then the first one found will be returned.
- Line 9 tells Express to render the update template, replacing variables with the given values as it goes. Important to note here is that we are also replacing values in the hidden form fields we created earlier with the current values of the customer details. This is to ensure that we update or delete the correct customer.

In your index.html file, add the following code after the </form> tag:

<br>
<form action="/get" method="GET">
  <input type="submit" value="Update/Delete">
</form>

This adds a new button that will make a GET request to /get, which will then return get.html.

Make a new file called get.html with the following contents:

<!DOCTYPE html>
<html  lang="en">
<body>
<form action="/get-client" method="GET">
  <label for="name">Customer name *</label><br>
  <input type="text" id="name" name="name" class="textInput" placeholder="John Smith" required>
  <input type="submit" value="Get customer">
</form>
</body>
</html>

This makes a simple form with an input for the customer's name and a button.

Clicking this button will make a GET request to /get-client, which will respond with the client details, and we will be able to update or delete them.

To see the customer details on a form after requesting them, we need a templating engine to render them onto the HTML page and send the rendered page back to us. With a templating engine, you define a template - a page with variables in it - and then give it the values you want to fill into the variables. In our case, we're going to request the customer details from the database and tell the templating engine to render them onto the page.

We're going to use a templating engine called Pug. Pug is a simple templating engine that integrates fully with Express. The syntax that Pug uses is very similar to HTML. One important difference in the syntax is that spacing is very important as it determines your parent/child hierarchy.

First, we need to tell Express which templating engine to use and where to find our templates. Put the following line above your route definitions (i.e. after the other app.use lines in index.js):

app.engine('pug', require('pug').__express)
app.set('views', '.')
app.set('view engine', 'pug')

Now create a new file called update.pug with the following content:

html
  body
    p #{message}
    h2= 'Customer details'
    form(method='POST' action='/update')
      input(type='hidden' id='oldname' name='oldname' value=oldname)
      input(type='hidden' id='oldaddress' name='oldaddress' value=oldaddress)
      input(type='hidden' id='oldtelephone' name='oldtelephone' value=oldtelephone)
      input(type='hidden' id='oldnote' name='oldnote' value=oldnote)
      label(for='name') Customer name:
      br
      input(type='text', placeholder='John Smith' name='name' value=name)
      br
      label(for='address') Customer address:
      br
      input(type='text', placeholder='42 Wallaby Way, Sydney' name='address' value=address)
      br
      label(for='telephone') Customer telephone:
      br
      input(type='text', placeholder='+275554202' name='telephone' value=telephone)
      br
      label(for='note') Customer note:
      br
      input(type='text', placeholder='Likes unicorns' name='note' value=note)
      br
      button(type='submit' formaction="/update") Update
      button(type='submit' formaction="/delete") Delete

This is very similar to the HTML form we created previously for create.html, however this is written in the Pug templating language. We're creating a hidden element to store the "old" name, telephone, address, and note of the customer - this is for when we want to do an update.

Using the old details to update the customer is an easy solution, but not the best solution as it makes the query cumbersome and slow. If you add extra fields into your database, you would have to remember to update your query as well, otherwise it could lead to updating or deleting the wrong customer if they have the same information. A better but more complicated way is to use the unique ID of the database document, as that will only ever refer to one customer.

We have also put in placeholder variables for name, address, telephone, and note, and we have given the form two buttons with different actions.

If you now run the code, you will have an index page with two buttons. Pressing the "Update/Delete" button will take you to a new page that asks for a customer name. If you fill in the customer name and press "Get customer", a page will load with the customer's details and two buttons below, "Update" and "Delete". Make sure you enter a customer name you have entered before.

Our next step is to add the "Update" and "Delete" functionality. Add the following code below your routes in index.js:

app.post('/update', function(req, res) {
  client.connect(err => {
    if (err) throw err;
    let query = { name: req.body.oldname, address: req.body.oldaddress, telephone: req.body.oldtelephone, note: req.body.oldnote };
    let newvalues = { $set: {name: req.body.name, address: req.body.address, telephone: req.body.telephone, note: req.body.note } };
    client.db("crmdb").collection("customers").updateOne(query, newvalues, function(err, result) {
        if (err) throw err;
        console.log("1 document updated");
        res.render('update', {message: 'Customer updated!', oldname: req.body.name, oldaddress: req.body.address, oldtelephone: req.body.telephone, oldnote: req.body.note, name: req.body.name, address: req.body.address, telephone: req.body.telephone, note: req.body.note});
      });
  });
})

app.post('/delete', function(req, res) {
  client.connect(err => {
    if (err) throw err;
    let query = { name: req.body.name, address: req.body.address ? req.body.address : null, telephone: req.body.telephone ? req.body.telephone : null, note: req.body.note ? req.body.note : null };
    client.db("crmdb").collection("customers").deleteOne(query, function(err, obj) {
      if (err) throw err;
      console.log("1 document deleted");
      res.send(`Customer ${req.body.name} deleted`);
    });
  });
})

This introduces two new POST handlers - one for /update, and one for /delete.

Line 2 connects to our MongoDB database.
Line 3 throws an error if there was a problem connecting to the database.
Line 4 defines a query that we will use to find the document to update. In this case, we'll use the details of the customer before it was updated. We saved this name earlier in a hidden field in the HTML. Trying to find the customer by its updated name obviously won't work, because it hasn't been updated yet. Also, note that we are setting some of the fields to null if they are empty. This is so that the database returns the correct document when we update or delete - if we search for a document that has no address with an address of '' (empty string), then our query won't return anything.
Line 5 defines the new values that we want to update our customer with.
Line 6 updates the customer with the new values using the query.
Line 7 throws an error if there was a problem with the update.
Line 8 logs that a document was updated.
Line 9 re-renders the update page with a message saying that the customer was updated, and displays the new values.
Line 15 connects to our MongoDB database.
Line 16 throws an error if there was a problem connecting to the database.
Line 17 defines a query that we will use to find the document to delete. In this case, we use all the details of the customer before any changes were made on the form to make sure we delete the correct customer.
In Line 18, we connect to the database and delete the customer.
Line 19 throws an error if there was a problem with the delete.
Line 20 logs that a document was deleted.
Line 21 sends a response to say that the customer was deleted.

Putting it all together

If you run your application now, you'll be able to create, update, and delete documents in a MongoDB database. This is a very basic CRUD application, with a very basic and unstyled UI, but it should give you the foundation to build much more sophisticated applications.

Some ideas for this are:

You could add fields to the database to classify customers according to which stage they are in your sales pipeline so that you can track if a customer is potentially stuck somewhere and contact them to re-engage.
You could integrate some basic marketing automation with a page allowing you to send an email or SMS to customers (though don't spam clients!).
You could also add fields to keep track of customer purchasing information so that you can see which products do well with which customers.

If you want to start from where this tutorial leaves off, fork the repl at https://replit.com/@ritza/replcrm.
To get additional guidance from the Replit community, you can join Replit's Discord server by using this invite link https://replit.com/discord.

https://replit.com/@ritza/replcrm

Building a Discord bot with Node.js and Replit

RitzaCo — Mon, 12 Sep 2022 08:31:21 +0000

In this tutorial, we'll use Replit and Node.js to build a Discord chatbot. The bot will be able to join a Discord server and respond to messages.

If you prefer Python, here's a Python Discord bot tutorial.

You'll find it easier to follow along if you have some JavaScript knowledge and you should have used Discord or a similar app such as Skype or Telegram before. We won't be covering the very basics of Node.js, but we will explain each line of code in detail, so if you have any experience with programming, you should be able to follow along.

Overview and requirements

We'll be doing all of our coding through the Replit web IDE, and we'll host our bot with Replit too, so you won't need to install any additional software on your machine. For this tutorial you will need to create a Discord account (if you already have one, you can skip this).

We will cover:

Creating an application and a bot user in your Discord account.
Creating a server on Discord.
Adding our bot to our Discord server.

Let's get through these admin steps first and then we can get to the fun part of coding our bot.

Creating a bot in Discord and getting a token

You can sign up for a free account over at Discord, and you can download one of their desktop or mobile applications from the Discord homepage. You can also use Discord in the browser.

Once you have an account, you'll want to create a Discord application. Visit the Discord developer's page and press the "New application" button, as in the image below.

Fill out a name for your bot and select "Create".

The first thing to do on the next page is to note your Application ID, which you'll need to add the bot to the server. You can come back later and get it from this page, or copy it somewhere so you can find it when you need it.

You can also rename the application and provide a description for your bot, then press "Save Changes".

You have now created a Discord application. The next step is to add a bot to this application, so head over to the "Bot" tab using the menu on the left and press the "Add Bot" button, as shown below. Click "Yes, do it" when Discord asks if you're sure about bringing a new bot to life.

The last thing we'll need from our bot is a token. Anyone who has the bot's token can prove that they own the bot, so you'll need to be careful not to share this with anyone. You can get the token by pressing "Reset Token", and then copy it to your clipboard by pressing "Copy".

Take note of your token or copy it to your clipboard, as we'll need to add it to our code soon.

Creating a Discord server

If you don't have a Discord server to add your bot to, you can create one by either opening the desktop Discord application that you downloaded earlier or returning to the Discord home page in your browser. Press the "+" icon, as shown below, to create a server.

Press "Create a server" in the screen that follows, and then give your server a name. Once the server is up and running, you can chat with yourself, or invite some friends to chat with you. Soon we'll invite our bot to chat with us as well.

Adding your Discord bot to your Discord server

Our Discord bot is still just a shell at this stage as we haven't written any code to allow it to do anything, but let's go ahead and add the bot to our Discord server anyway. To add a bot to your server, you'll need the Application ID from the "General Information" page that we looked at before when we created our ReplBotApplication (ie. the application ID, not the secret bot token).

Create a URL that looks as follows, but using your application ID instead of mine at the end (the link calls the application ID "client_id"):

https://discordapp.com/api/oauth2/authorize?scope=bot&client_id=746269162917331028

Visit the URL that you created in your web browser and you'll see a page similar to the following where you can choose which server to add your bot to.

Select the server we created in the step before this and hit the "Authorize" button. After completing the captcha, you should get an in-app Discord notification telling you that your bot has joined your server.

Now we can get to the fun part of building a brain for our bot!

Creating a repl and installing our Discord dependencies

The first thing we need to do is create a Node.js repl to write the code for our Discord bot. Over at Replit, create a new repl, choosing "Node.js" as your language.

We don't need to reinvent the wheel as there is already a great Node wrapper for the Discord bot API called discord.js. Normally we would install this third-party library through npm, but because we're using Replit, we can skip the installation. Our repl will automatically pull in all dependencies.

In the default index.js file that is included with your new repl, add the following line of code:

const Discord = require('discord.js');

Press the "Run" button and you should see Replit installing the Discord library in the output pane on the right, as in the image below.

Our bot is nearly ready to go - but we still need to plug in our secret token. This will authorize our code to control our bot.

Setting up authorization for our bot

By default, Replit code is public. This is great as it encourages collaboration and learning, but we need to be careful not to share our secret bot token (which gives anyone who has access to it full control of our bot).

To get around the problem of needing to give our code access to the token while allowing others to access our code but not our token, we'll be using environment variables. Replit allows us to set secret environment variables through the "Secrets (Environment variables)" menu option.

Open the "Secrets" menu option. There you will be able to set environment variables for your repl. Set the key as the name of your environment variable to DISCORD_BOT_SECRET. Set the value as your bot's secret token (note that this is the second token that we got while setting up the bot -- different from the Application ID that we used to add our bot to our server). It should look something like:

Let's make a Discord bot that repeats everything we say but in reverse. We can do this in just a few lines of code. In your index.js file, add the following:

const Discord = require("discord.js");
const client = new Discord.Client({
  intents: [
    "GUILDS",
    "GUILD_MESSAGES",
  ],
});

const token = process.env['DISCORD_BOT_SECRET']

client.on('ready', () => {
  console.log("I'm in");
  console.log(client.user.username);
});

client.on('messageCreate', msg => {
    if (msg.author.id != client.user.id) {
        msg.channel.send(msg.content.split('').reverse().join(''));
    }
});

client.login(token)

Let's tear this apart line by line to see what it does.

Line 1 is the line we added earlier. This line tells Replit to install the third party library and brings it into this file so that we can use it.
In line 2, we create a Discord Client. We'll use this client to send commands to the Discord server to control our bot and send it commands.
In line 3-8, we provide the intents of our Client, these are provided to designate which events our bot will be able to receive.
In line 9, we retrieve our secret token from the environment variables (which Replit sets from the "Secrets" menu).
In line 11, we define an event for our client, which defines how our bot should react to the ready event. The Discord bot is going to run asynchronously, which might be a bit confusing if you're used to running standard synchronous code. We won't go into asynchronous coding in-depth here, but if you're interested in what this is and why it's used, there's a good guide over at RisingStack. In short, instead of running the code in our file from top to bottom, we'll be running pieces of code in response to specific events.
In lines 12-14, we define how our bot should respond to the ready event, which is fired when our bot successfully joins a server. We instruct our bot to output some information server-side (i.e. it will be displayed in our repl's output, but not sent as a message through to Discord). We'll print a simple "I'm in" message to see that the bot is there and print our bot's username (if you're running multiple bots, this will make it easier to work out who's doing what).
Lines 16-20 are similar, but instead of responding to a ready event, we tell our bot how to handle new messages. Line 17 says we only want to respond to messages that aren't from us (otherwise our bot will keep responding to himself -- you can remove this line to see why that's a problem), and line 18 says we'll send a new message to the same channel we received a message in (msg.channel) and the content we'll send will be the same message that we received, but backwards. To reverse a string, we split it into its individual characters, reverse the resulting array, and then join it all back into a string again.

The last line fires up our bot and uses the token we loaded earlier to log into Discord.

Press the "Run" button again and you should see your bot reporting a successful channel join in the repl output.

Open Discord, and from within the server we created earlier, you will be able to send a message (by typing into the box highlighted below) and see your bot respond!

The bot responds each time, reversing the text we enter.

Keeping our bot alive

Your bot can now respond to messages, but only for as long as your repl is running. If you close your browser tab or shut down your computer, your bot will stop and no longer respond to messages on Discord.

Replit will keep your code running after you close the browser tab only if you are running a web server. Our bot doesn't require an explicit web server to run, but we can create a server and run it in the background just to keep our repl alive.

Create a new file in your project called keep_alive.js and add the following code:

var http = require('http');

http.createServer(function (req, res) {
  res.write("I'm alive");
  res.end();
}).listen(8080);

We won't go over this in detail as it's not central to our bot, but here we start a web server that will return "I'm alive" if anyone visits it.

In our index.js file, we need to add a require statement for this server at the top. Add the following line near the top of index.js:

const keep_alive = require('./keep_alive.js')

After doing this and hitting the green "Run" button again, you should see some changes to your repl. For one, you'll see a new pane in the top right that shows the web output from your server. We can see that visiting our repl now returns a basic web page showing the "I'm alive" string that we told our web server to return by default.

Now your bot will stay alive even after closing your browser or shutting down your development machine. Replit will still clean up your server and kill your bot after about one hour of inactivity, so if you don't use your bot for a while, you'll have to log into your repl and start the bot up again. Alternatively, you can set up a third-party (free!) service like Uptime Robot. Uptime Robot pings your site every five minutes to make sure it's still working -- usually to notify you of unexpected downtime, but in this case the constant pings have the side effect of keeping our repl alive as it will never go more than an hour without receiving any activity. Note that you need to select the "HTTP" option instead of the "Ping" option when setting up Uptime Robot, as Replit requires regular HTTP requests to keep your chatbot alive, or you can use an Always-on repl to keep it running 24/7.

Forking and extending our basic bot

This is not a very useful bot as-is, but the possibilities are only limited by your creativity now! You can have your bot receive input from a user, process the input, and respond in any way you choose. In fact, with the basic input and output that we've demonstrated, we have most of the components of any modern computer, all of which are based on the Von Neumann architecture (we could easily add the missing memory by having our bot write to a file, or with a bit more effort link in a SQLite database for persistent storage).

If you followed along with this tutorial, you'll have your own basic repl bot to play around with and extend. If you were only reading, you can easily fork my bot at https://replit.com/@ritza/discord-bot-node-1 and extend it however you like (you'll need to add your own token and reset the secrets still). Happy hacking!

If you're stuck for ideas, you could try link up your Discord bot to the Twitch API to get notified when your favorite streamers are online, or build a text adventure.