my apps were invisible to AI agents — here's what I am doing about it

Jill Mercer — Mon, 18 May 2026 01:22:32 +0000

i'm an indie app builder and vibe coder. i've shipped over 30 small business apps — invoicing, inventory, packing slips, tax tracking. and now apparently an open standard for ai agents.

that last one surprised me too.

the problem i kept running into: even the best ai agents hallucinate when they look at a web app. they guess where to click. they call the wrong tools. they fail silently and confidently. i had 30 apps that existed — agents just couldn't see them.

so i built blueprint protocol. a blueprint.txt file at the root of your app that tells agents exactly what it can do, which tools to call, what's human-only, and how to complete each flow step by step. no mcp required. it works for any agent trying to navigate any web app.

then i ran a benchmark and found out it also cuts mcp discovery overhead by 78%.

what it does for mcp specifically
mcp has two calls. tools/list returns every tool the server exposes. tools/call calls a specific tool by name.

without a blueprint, an agent calls tools/list first, reads through all N tool definitions, picks one, then calls it.

with a blueprint, the agent already knows the tool name and exact parameters. it calls tools/call directly. tools/list never happens.

that's not just cheaper discovery. it's the removal of an entire round trip.

the benchmark
i ran a controlled test across four models — claude, gpt-4o, gemini, and grok — five candidate mcp servers, one correct match. agents were told to find the best server efficiently. no blueprint references in the harness instructions. discovery was organic.

model without blueprint with blueprint reduction
claude 58 pts 13 pts ~78%
gpt-4o 58 pts 28 pts ~54%
gemini 23 pts 9 pts ~61%
grok 58 pts 58 pts 0%
overhead scored by weighted resource inspection cost — llms.txt at 10pts, blueprint.txt at 1pt, reflecting typical real-world document sizes.

grok's 0% is an honest result. not all models follow discovery signals today. that's expected — robots.txt and llms.txt didn't work everywhere on day one either.

v3.0.0 — agent-first structure
the spec just hit v3.0.0. two changes that matter for mcp:

[MCP] flag on line one — agents confirm server availability instantly
format b capabilities index — root file is ~150 tokens, agents fetch only the one capability file they need (~350 tokens). total: ~500 tokens vs thousands from tools/list
try it
spec: https://github.com/Explorer-64/blueprint-protocol

live reference implementation: https://imagcon.app/blueprint.txt

benchmark results: https://stackapps.app/mcp-blueprint-results

open standard. mit licensed. no tooling required. works with or without mcp. if you have a web app or an mcp server, a blueprint takes about 10 minutes.

I shipped 30 apps. AI crawlers ignore them.

Jill Mercer — Sun, 10 May 2026 16:43:48 +0000

I have shipped over 30 apps. Most of them are absolute ghosts right now.

If you ask an AI tool to find something that does exactly what my projects do, they won't show up. It is not because the apps are bad. It is because I built them before AI crawlers mattered.

I learned the hard way that the ground moves fast in this industry. I am self-taught — learned by shipping, not by studying. A while back, I started vibe coding with Bolt. Then I found Databutton. I loved it — the community was real, and it is where I actually learned how to vibe code. Then they pivoted to B2B overnight. They shut down the Discord. Pricing shot up. The little guy got pushed out.

I was weeks away from monetizing. Instead, I had to pack up and leave. I jumped to Replit, tried Firebase Studio, messed with Google AI Studio, and finally landed on Cursor. Moving those apps one by one took forever. Six months just evaporated. That whole mess taught me a hard lesson: own your stack. Own your presentation layer. Never let a platform sit between you and your users.

So now I do. I build everything in Cursor. I own the code.

But owning the stack does not solve the discovery problem. People are searching differently now. They ask Perplexity, ChatGPT, or Claude for tool recommendations instead of scrolling Google. And those AI bots do not read websites the way humans do.

Here is what I have learned from watching my older apps fail to rank in AI search:

1. Client-side rendering is a wall.
Most of my old projects are pure client-side React. To a human, they load fast and look great. To an AI crawler that does not execute JavaScript, they look like a blank page with a <div id="root">. If the bot cannot read the text in the initial HTML, it just leaves.

2. Words matter more than pictures.
I used to rely on slick screenshots and minimal copy for landing pages. AI vision models might be smart, but standard web crawlers still run on plain text. I have started rewriting my landing pages to clearly explain the problem and the solution in plain English.

3. Semantic HTML is back.
Div soup is out. Using actual <header>, <nav>, <main>, and <article> tags helps the AI understand the hierarchy of the page. It needs to know what is navigation and what is the actual product description.

4. Structured data helps.
I used to ignore JSON-LD and schema markup. It felt like boring SEO work. Now, I put it everywhere. It feeds the bots exact data about what the app is, who made it, and what it costs.

I am still figuring this out. I am currently working through the discovery problem one app at a time to make my work visible to LLMs. It is frustrating to realize the rules changed while I was busy migrating code — but that is the job.

We spent a decade optimizing for Google search. Now we have to figure out how to talk to the bots.

How are you handling this shift? Are you doing anything specific to get your projects seen by AI search tools?

DEV Community: Jill Mercer

my apps were invisible to AI agents — here's what I am doing about it

I shipped 30 apps. AI crawlers ignore them.