DEV Community: Ahmed Boateng

Can Claude Connect to LinkedIn? What Works in 2026

Ahmed Boateng — Mon, 19 Oct 2026 23:30:41 +0000

The short answer. Yes, through other tools. LinkedIn hasn't published an MCP server or a Claude connector, and its official APIs mostly need LinkedIn's approval. Claude still reaches LinkedIn four ways: posting and analytics tools like Taplio, outreach tools that run your own account like HeyReach, data services that read public posts and the people who engaged with them, and Claude in Chrome driving your own browser. Pick by the job, and know LinkedIn's limits on automation before you do.

Can Claude connect to LinkedIn?

Yes, but not through LinkedIn itself. As of 12 October 2026, LinkedIn has not published an MCP (Model Context Protocol) server, and Claude's connector directory lists no connector made by LinkedIn. What it does list are connectors made by other companies that work with LinkedIn. Several are marked as verified by Anthropic, and any tool with an MCP server can be added as a custom connector.

So the useful question is what you want Claude to do on LinkedIn. There are four jobs, and each has its own route.

Job	Route	Uses your LinkedIn account?
Post, schedule, read your analytics	Posting tools with a Claude connector (Taplio)	Yes, with your permission
Send connection requests and messages	Outreach tools with an MCP server (HeyReach, La Growth Machine, Waalaxy, Expandi, PhantomBuster)	Yes, they run your account
Find leads in public posts and their engagers	Data services that read public posts (Layrcake, Apify actors)	No
Build an app on LinkedIn data	LinkedIn's official APIs	Your app, with LinkedIn's approval

Checked 12 October 2026 against each company's own pages.

Does LinkedIn have an official API or MCP server?

LinkedIn has official APIs but no MCP server. According to LinkedIn's developer documentation, only two permissions are open to every developer without approval: Sign In with LinkedIn, and Share on LinkedIn (posting, commenting and liking for the person who signed in).

The marketing APIs, Community Management included, are open to approved developers through an application, and the development tier allows 500 calls per app per day (Increasing Access). The Sales Navigator API is not taking new partners: its page says "We are not currently accepting new partners" (Sales Navigator Application Platform). Nothing in the open permissions lets an app search other members or send messages.

That gap is why a whole category of third-party tools exists.

Can Claude send LinkedIn messages and connection requests?

Yes, through outreach tools that run your own LinkedIn account and have their own MCP server. You connect the tool to Claude, and Claude asks it to build a campaign, add people or draft messages. The tool does the sending from your account, inside the daily limits it sets.

Tool	What its own docs say Claude can do	Plan for MCP access
HeyReach	Build and launch campaigns, segment lead lists, draft messages, read reply sentiment	Growth plan ($79 a month per sender) and up; the MCP itself is free
La Growth Machine	Build audiences from a Sales Navigator search, manage campaigns, send LinkedIn messages	Active subscription; 14-day trial
Waalaxy	List prospect lists and campaigns, add prospects to a campaign	Advanced and Business plans
Expandi	Qualify leads and build campaign drafts; campaigns stay drafts until you launch	Active subscription or 7-day trial
PhantomBuster	Launch and monitor automations, fetch their results, build lead lists	Not stated on its MCP page

From each vendor's MCP or help pages, 12 October 2026. Plans and prices change; check before you buy.

Limits matter more than tools here. HeyReach, for example, caps each sender at 200 connection requests a week and recommends 25 a day once an account is warmed up (HeyReach help). Our view: start below whatever cap your tool sets, and let Claude draft while you approve the first batches by hand.

Does LinkedIn allow automation?

LinkedIn's User Agreement (section 8.2, effective 3 November 2025) says members must not use "bots or other unauthorized automated methods to access the Services". It also rules out software that scrapes or copies profiles. Its help centre names browser extensions that "automate activity on LinkedIn's website" among the tools it doesn't permit, and warns that accounts using them can be restricted (Prohibited software and extensions).

Outreach tools are used widely anyway, and most publish conservative daily limits to keep accounts safe. The risk sits with the account that sends. Weigh it the way you would for any account you can't easily replace.

Can Claude in Chrome use LinkedIn for me?

It can operate a LinkedIn tab you are signed in to, the same way it operates any site. Claude in Chrome is a browser extension on Claude's paid plans that can "read, click, and navigate websites alongside you" (Anthropic).

Anthropic's own safety guidance puts the site's rules on you: users are responsible for "respecting third-party website terms of service, including any restrictions on automated access" (Use Claude in Chrome safely). For LinkedIn, that means the 8.2 limits above. It works well for one-off help, like reading a profile with you. We wouldn't use it to run outreach.

How do you find leads on LinkedIn with Claude without using your account?

Read public posts and the people who engaged with them, through a data service, and keep your account out of it. Someone who commented on a post about your problem this week is a better lead than a name from a list. Claude does the filtering.

We use Layrcake for this. Claude asks for a person's or company's recent posts, then for everyone who reacted or commented on one of them. Layrcake reads public data, so no LinkedIn login or session cookie of yours is involved. Each person comes back with a profile link, their headline and what they did on the post.

The job we gave Claude

Find this person's most-liked LinkedIn post from the last 30 days, read everyone who engaged with it, and keep only heads of sales, VPs of sales and CROs. Tell me what it will cost before reading everyone.

In our run on 3 October, Claude quoted the cost first, then read all 563 people who engaged with one post in the background, in about two and a half minutes. It kept 26 sales leaders, for 232.2 credits. Six of the 26 had titles a keyword match would have missed, like Chief Commercial Officer, and were added by a title check.

From there Claude can find work emails for the people you keep and draft a first message, which you send from your own inbox or LinkedIn by hand. Apify's post-reaction and post-comment scrapers are another route to the same public data, priced per thousand results and reachable through the Apify connector in Claude's directory.

Posting is a separate job. Reading public posts tells you who to talk to. To publish posts or read your own analytics from Claude, use a posting tool such as Taplio, which has a connector in Claude's directory.

Which route should you pick?

Pick by the job, not the tool. To post and track your own content, use a posting connector. To send LinkedIn messages at volume, use an outreach tool with an MCP server, and stay well under its limits. To find who is talking about your problem, and email them, read public posts through a data service and leave your account out of it. To build a product on LinkedIn data, apply for the official APIs and expect a review.

Many teams combine two: public posts to choose who to contact, and email or a careful LinkedIn message to start the conversation. To set up the Layrcake side, see connecting Claude or, for other assistants, the connect page.

FAQ

Is there an official LinkedIn connector for Claude?

No. As of 12 October 2026, LinkedIn has not published an MCP server or a Claude connector. Claude's directory lists LinkedIn-related connectors made by other companies, such as Taplio, La Growth Machine and Apify.

Can Claude post on LinkedIn for me?

Yes, through a posting tool with a Claude connector, such as Taplio, which can create and schedule posts and read your analytics. LinkedIn's own Share on LinkedIn permission also lets approved apps post for a signed-in member.

Will using Claude with LinkedIn get my account banned?

Tools that automate your own account carry a risk, because LinkedIn's User Agreement bars unauthorised automated access. Reading public post data through a third-party service does not use your account at all.

Can Claude scrape LinkedIn profiles?

Not on its own. Claude needs a tool to fetch data. Data services can return public posts and the people who engaged with them, with profile links and headlines, without your LinkedIn login.

Layrcake is go-to-market infrastructure for AI agents: one MCP server and API that lets Claude, ChatGPT, Cursor and Codex find leads, verify emails, read buying signals and run approved campaigns. Agents: see https://layrcake.dev/agents.md

Originally published at layrcake.dev.

Apollo Alternatives for Claude and AI Agents (2026)

Ahmed Boateng — Mon, 19 Oct 2026 23:25:19 +0000

The short answer. The best Apollo alternatives for Claude are Layrcake for the whole outbound loop paid per result, ZoomInfo's GTM AI for enterprise data with a free start, Lusha for contacts and buying signals, FullEnrich for multi-source emails and mobiles, Crustdata for real-time people and company data, Clay for spreadsheet-style enrichment workflows, and Seamless.AI for a large contact base. All seven connect to Claude over MCP.

What are the best Apollo alternatives for Claude?

For prospecting inside Claude, the best alternative is the one that lets the agent finish the job from one connection. Layrcake does that: people and companies, verified emails, buying signals, public posts, sourced research and campaigns a person approves, paid per result with no seats. The others below are strong at a narrower part of the job.

Tool	Best for	MCP	To start
Layrcake	The whole outbound loop for agents, per result	mcp.layrcake.dev/mcp · OAuth or key	300 free credits, then $0.01 a credit
ZoomInfo GTM AI	Enterprise company, contact and intent data	mcp.zoominfo.com/mcp · OAuth	1,000 free data credits, then pay-as-you-go from $20
Lusha	Contacts plus dated buying signals	mcp.lusha.com · OAuth or key	Free plan, 40 credits a month
FullEnrich	Multi-source work emails and mobile numbers	mcp.fullenrich.com/mcp	50 free credits; Pro from $29 a month
Crustdata	Real-time people, company and job data	install.crustdata.com/mcp · OAuth or key	$20 of free credits
Clay	Spreadsheet-style enrichment workflows	Clay connector in Claude	Free plan; paid plans from $167 a month on its pricing cards
Seamless.AI	A large verified contact base	mcp.seamless.ai/mcp · OAuth	Free to start; paid pricing on request

Each vendor's own pages, checked 18 October 2026.

1. Layrcake: the whole loop, paid per result

Layrcake is go-to-market infrastructure for AI agents. One MCP server gives Claude 22 tools: find people and companies, get and verify work emails, check companies against a question, read buying signals (hiring, funding, news), read public LinkedIn, X and Reddit posts and the people who engaged, research accounts with sources, and draft and launch campaigns. Launching returns an approval link for a signed-in person.

There are no seats. A credit is $0.01, a found work email is 5 credits, a verification 1.25, a lookup that finds nothing is free, and credits never expire. In our run on 1 October 2026, Claude took 20 heads of sales to 15 verified, ready-to-send drafts for 145 credits. See Layrcake vs Apollo for the side-by-side.

2. ZoomInfo GTM AI: enterprise data, now self-serve

ZoomInfo's GTM AI offers its company, contact and intent data through one API and an MCP server, free to start with 1,000 data credits and 1,000 AI credits, then pay-as-you-go from $20. Search, lookup and find-similar are free; credits go on enriched records and AI research (GTM AI pricing). It suits teams who want ZoomInfo's data inside Claude. Layrcake vs ZoomInfo compares the two.

3. Lusha: contacts and buying signals

Lusha calls itself "the B2B data and intelligence layer for GTM teams and AI agents", with contacts, company data and dated buying signals in Claude, ChatGPT and any MCP client. Its MCP server lists 50 tools and is in Claude's connector directory. Plans start free with 40 credits a month; API access comes with Pro (Lusha pricing).

4. FullEnrich: emails and mobiles from many sources

FullEnrich runs a waterfall across "25+ data sources" for work emails, personal emails and mobile numbers, and says "A lookup that returns nothing costs nothing". Its MCP is included on every plan, from a 50-credit free trial to Pro from $29 a month (FullEnrich pricing). A good fit when phone numbers matter most.

5. Crustdata: real-time data for agents

Crustdata describes itself as a "Real-Time B2B Data Provider for AI Agents", with people, company, job and post data and watchers. New accounts get $20 of credits, a search that comes back empty costs nothing, and real-time API access starts at four figures a month (Crustdata pricing). See Layrcake vs Crustdata.

6. Clay: enrichment workflows in a table

Clay calls itself "Infrastructure to get any data, run agentic workflows, and launch GTM plays", with 150-plus data providers in a table and a connector in Claude. Failed enrichments are not charged. Its pricing cards on 18 October 2026 show a free plan and paid plans from $167 a month (Clay pricing). We compared working in Claude with working in Clay in Claude vs Clay.

7. Seamless.AI: a large contact base

Seamless.AI offers verified contacts and direct dials, with an MCP server of "54 tools across 11 domains, from search to send" that signs in with OAuth. Search is free and credits are used on enrichment; paid pricing is quoted on request (Seamless.AI MCP).

How do you choose?

Ask what the agent has to finish. If it is a list, any data tool above will do. If it is outreach (a checked list, a reason to write to each person, a sequence, a send a person approved and the replies), choose the tool that covers that in one connection and charges only for results. That is what we built Layrcake for. Connect it to Claude in about two minutes; new accounts get 300 free credits.

FAQ

What is the best Apollo alternative for Claude?

For agent-run prospecting, Layrcake: one MCP connection covers people, verified emails, buying signals, public posts, research and approved campaigns, paid per result with no seats and credits that never expire.

Is there a free Apollo alternative for AI agents?

Several start free: Layrcake with 300 credits, ZoomInfo's GTM AI with 1,000 data credits, Lusha with 40 credits a month, FullEnrich with 50 credits and Crustdata with $20 of credits.

Which Apollo alternatives have an MCP server?

All seven here: Layrcake, ZoomInfo GTM AI, Lusha, FullEnrich, Crustdata, Clay (through its Claude connector) and Seamless.AI.

Originally published at layrcake.dev.

Best Tools for Prospecting With Claude (2026)

Ahmed Boateng — Mon, 19 Oct 2026 23:19:57 +0000

The short answer. The best tool for prospecting with Claude is one connection that covers the whole job: Layrcake finds people and companies, gets and verifies emails, reads buying signals and public posts, researches accounts with sources and drafts campaigns a person approves. For single jobs: Apollo or ZoomInfo's GTM AI for contact databases, Lusha for signals, FullEnrich or Hunter for emails, Clay for table-based workflows, and Claude's own web search for open-web reading.

What are the best tools for prospecting with Claude?

The best setup is the fewest tools that cover the job, because every connection adds tools Claude has to choose between. Layrcake covers the whole prospecting job in one MCP connection. If you only need one part of it, the specialists below are good at their part.

Prospecting job	Best in one connection	Specialists
Find people and companies	Layrcake	Apollo, ZoomInfo GTM AI, Crustdata
Work emails and phones	Layrcake (emails, verified)	FullEnrich, Hunter, Lusha
Check an address before sending	Layrcake	Hunter
Buying signals (hiring, funding, news)	Layrcake	Lusha, ZoomInfo GTM AI
Who engaged with a LinkedIn post	Layrcake	PhantomBuster, Apify (both connect to Claude)
Account research with sources	Layrcake	ZoomInfo GTM AI, Claude's web search
Draft and send a campaign	Layrcake, with a person approving	Apollo, Instantly, lemlist

From each vendor's own pages, checked 18 October 2026.

1. Layrcake: every prospecting job in one connection

Connect https://mcp.layrcake.dev/mcp in Claude and ask for the job in one sentence:

Claude

Find 20 heads of sales at SaaS companies in London with 20 to 200 staff. Verify every email, find a recent reason to reach out to each one, and draft a three-step sequence that opens with it. Stop before sending.

In our run on 1 October 2026, Claude found 20 people, got 16 emails, kept 15 after a second check, found recent LinkedIn activity for all 15 and drafted the sequence: 145 credits ($1.45), with the launch waiting on an approval link. A lookup that finds nothing is free, and credits never expire. The full run has every tool call.

2–3. Contact databases: Apollo and ZoomInfo GTM AI

Apollo brings its contact database, sequences and tasks into Claude through a hosted MCP server, on plans from a free tier to $119 per seat a month billed annually (Apollo pricing). ZoomInfo's GTM AI brings company, contact and intent data, free to start with 1,000 data credits and pay-as-you-go from $20 (GTM AI pricing). Compare them with Layrcake: vs Apollo, vs ZoomInfo.

4–6. Emails and signals: FullEnrich, Hunter and Lusha

FullEnrich finds work emails and mobiles across 25-plus sources and charges nothing when it finds nothing (FullEnrich). Hunter finds and verifies emails, with its MCP included on all plans (Hunter). Lusha adds dated buying signals to contacts, with 50 MCP tools and a listing in Claude's directory (Lusha). For how finding and checking emails works, read our email guide.

7–8. Workflows and the open web: Clay and Claude's web search

Clay runs enrichment as table workflows across 150-plus providers and connects to Claude; see Claude vs Clay. Claude's own web search reads company websites and news without another tool, which is enough for a quick look at one account.

How do you pick?

Start from the job, then count connections. If Claude has to stitch three tools together for one list, its context fills with tool descriptions and its choices get worse. One connection that covers the job is faster, cheaper to run and easier to check. Connect Layrcake to Claude in about two minutes; new accounts get 300 free credits.

FAQ

What is the best tool for prospecting with Claude?

Layrcake covers the whole prospecting job in one MCP connection: people, verified emails, buying signals, public posts, sourced research and campaigns that a person approves, paid per result.

Can Claude prospect without any tools?

Only from what it already knows and the open web. Claude needs connected tools for live contact data, email addresses and verification.

How many tools should I connect to Claude for sales?

As few as cover the job. Each connection adds tool descriptions to Claude's context, so one connection that covers prospecting end to end usually works best.

Originally published at layrcake.dev.

Best MCP Servers for Sales and Go-To-Market Teams (2026)

Ahmed Boateng — Mon, 19 Oct 2026 23:14:35 +0000

The short answer. The best MCP servers for sales depend on the job. For the whole outbound loop in one connection, use an all-in-one like Layrcake, Apollo or Amplemarket. For contact data, Apollo, ZoomInfo, Crustdata, FullEnrich or Lusha. For sending, Instantly, Smartlead, lemlist or Outreach. For LinkedIn, HeyReach. Keep your CRM's own server (HubSpot, Salesforce, Attio) for records, and add Firecrawl, Exa or Perplexity for web research. Two or three well-chosen servers beat eight.

What are the best MCP servers for sales?

The best MCP servers for sales are the ones that cover your loop with the fewest connections: one for finding and checking people, one for sending, and your CRM's own server for the record. An MCP (Model Context Protocol) server lets an AI assistant such as Claude, ChatGPT or Cursor call a tool directly. Anthropic introduced the standard in November 2024, and by October 2026 almost every major sales tool has an official one.

We make Layrcake, one of the servers below, so read our entry with that in mind. Everything else here comes from each vendor's own documentation, read on 12 October 2026. Most "best MCP servers for sales" lists are written by a vendor that ranks itself first. We have grouped these by job instead, and said where someone else is the better pick.

Job	Servers	Pick when
The whole loop, one connection	Layrcake, Apollo, Amplemarket	You want find, check, send and replies without five accounts
Contact and company data	ZoomInfo, Crustdata, Clay, FullEnrich, Lusha	You need a specific dataset or enrichment depth
Email sending	Instantly, Smartlead, lemlist, Outreach	You already run sending on one of them
LinkedIn outreach	HeyReach	LinkedIn is your main channel
CRM	HubSpot, Salesforce, Attio	Always: keep the record where it lives
Web research	Firecrawl, Exa, Perplexity	Reading sites and the open web
Call intelligence	Gong	You want deal context from recorded calls

Which MCP servers cover the whole outbound loop?

Three servers cover finding people, checking them and sending, in one connection. They differ in where the data comes from and how you pay.

Layrcake. A hosted MCP server and REST API built for AI agents, at https://mcp.layrcake.dev/mcp. Claude and ChatGPT sign in with OAuth; Cursor, Codex and Claude Code can use a key. Its tools cover company and people search, work emails and phones, email verification, buying signals (hiring, funding, news), company research, public social posts and the people who engaged with them, campaigns and replies. One prepaid credit balance pays for all of it, lookups that find nothing are free, and every response reports its charge. Anything that sends email or spends money returns a preview and an approval link first. It does not replace a CRM; pair it with yours.

Apollo. A hosted server at https://mcp.apollo.io/mcp with OAuth, or an API key for headless clients. Apollo's docs list actions across people and company search, enrichment, lists, sequences, tasks and email sending. Its product page says any plan qualifies, including free; its docs add that free accounts on a personal email address can't use people search or enrichment (Apollo docs). A good pick if your team already lives in Apollo.

Amplemarket. A hosted server at https://mcp.amplemarket.com/mcp, OAuth only, launched in March 2026. Its tools span search, enrichment, lead lists, sequences, workflows, analytics and its unified inbox. Enrichment costs 0.5 Amplemarket credits (Amplemarket). A strong fit for teams buying a full sales platform.

Which MCP servers are best for lead data and enrichment?

Pick a data server for the dataset you trust and the way you want to pay.

Server	Connects	What it's for	To run it
ZoomInfo GTM AI	Hosted, OAuth	Company, contact and intent data, plus AI research	Free to start (1,000 data credits), then pay-as-you-go from $20
Crustdata	Hosted, OAuth	Real-time people, company, jobs and posts data; 23 tools	Credits per call (1–4 per company enrichment)
Clay	Built-in connector in Claude, ChatGPT and Codex	150+ data providers and your team's Clay workflows	500 free credits on first connect
FullEnrich	Hosted, OAuth	Multi-source work emails and mobile numbers	Search preview free; 0.25 credits per export
Lusha	Hosted, OAuth or key	Contacts, enrichment and fit scores	Any Lusha plan; calls use credits

From each vendor's MCP docs, 12 October 2026.

ZoomInfo's GTM AI now starts free, so its data is easy to try in Claude. If your RevOps team builds workflows in Clay, its connector lets reps run them from chat. FullEnrich is a good pick when finding mobile numbers is the main job.

Which MCP servers can send cold email?

Sending servers let the assistant build campaigns, add leads and read results in the tool that does the sending.

Server	Connects	What its docs list	To run it
Instantly	Hosted, API key	31 tools: campaigns, leads, email, analytics, accounts	Free with an Instantly subscription
Smartlead	Hosted (SSE only for now), API key	Campaign status, lead data, deliverability checks	Not stated
lemlist	Hosted, OAuth or key	Campaigns, leads, a 450M+ contact database, enrichment, verification	Some operations use credits
Outreach	Hosted, OAuth 2.1	32 tools: sequences, tasks, accounts, prospects, opportunities	A licensed seat with the Amplify add-on

From each vendor's MCP docs, 12 October 2026.

If your team already runs deliverability in Instantly or Smartlead, their own servers expose the mailbox and warm-up detail you'll want. Outreach's server is for enterprise teams on Outreach today; its support page notes that record updates aren't available yet.

Ask for an approval step. Whichever server sends, make sure the assistant stops at a draft or a preview and a person approves the launch. One wrong list sent at 2 a.m. costs more than any tool.

Is there an MCP server for LinkedIn?

LinkedIn hasn't published one. Outreach tools that run your own LinkedIn account have, and HeyReach is the most established: each workspace gets its own MCP URL and key, and it is built for running many sender accounts at once (HeyReach). La Growth Machine, Waalaxy, Expandi and PhantomBuster have servers too.

To find leads in public LinkedIn posts without using your account, a data server that reads posts and the people who engaged with them does that job; Layrcake and Crustdata both cover posts. Our LinkedIn guide compares every route.

Which CRM MCP servers should you connect?

Your own CRM's, always. It owns the record and the permissions, so let every other server feed it.

HubSpot's remote server at https://mcp.hubspot.com became generally available on 13 April 2026, with OAuth 2.1. It reads most CRM and marketing objects and writes contacts, companies, deals, tickets, products and activities (HubSpot changelog). Salesforce's hosted MCP servers reached general availability in April 2026 and connect through an External Client App your admin sets up (Salesforce). Attio's server at https://mcp.attio.com/mcp uses OAuth and covers records, lists, notes, tasks and reports (Attio).

Which MCP servers help with account research?

For reading company websites and the open web, add a research server.

Server	Connects	Good for
Firecrawl	Hosted, key or OAuth; keyless within daily limits	Turning websites into clean text: scrape, crawl, search
Exa	Hosted, keyless, OAuth or key	Web search built for AI, plus multi-step research runs
Perplexity	Hosted or local, OAuth or key	Search, answers with sources, deep research; standard API pricing
Gong	Official server, any Gong plan	Questions about accounts and deals from your recorded calls

From each vendor's MCP docs, 12 October 2026.

Data servers with company research built in, Layrcake and Clay among them, return researched facts with sources directly, which saves the assistant reading pages itself.

How do you choose an MCP server for your sales team?

Start from the job you do most, then check four things on the vendor's own MCP page.

How it connects. Hosted servers with OAuth are the easiest: paste a URL, sign in. Servers that take an API key in the URL need more care, because the key travels with the configuration.

What plan it needs. Some come free with any account; others need a licence, a seat or an add-on.

How it charges. Look for per-call prices and whether a lookup that finds nothing costs anything.

What it does before acting. Anything that sends email or spends money should show you a preview first.

Then keep the set small. Every server adds tools the assistant has to choose between, and two or three that cover your loop work better than eight that overlap. To try Layrcake's, see the MCP page or connect it to your assistant.

FAQ

What is an MCP server for sales?

A server that exposes a sales tool, such as a contact database, an email sender or a CRM, to an AI assistant through the Model Context Protocol. Once connected, Claude, ChatGPT or Cursor can call it directly, for example to search for leads or create a campaign.

Does Apollo have an MCP server?

Yes. Apollo runs a hosted MCP server at mcp.apollo.io with OAuth or an API key, and its docs list actions across search, enrichment, sequences and email.

Does HubSpot have an MCP server?

Yes. HubSpot's remote MCP server became generally available on 13 April 2026. It uses OAuth 2.1, reads most CRM objects and can write contacts, companies, deals, tickets, products and activities.

Which AI assistants support remote MCP servers?

Claude (custom connectors on every plan, one on Free), ChatGPT (custom MCP servers in developer mode), Cursor and VS Code all connect to remote MCP servers, many with OAuth sign-in.

Originally published at layrcake.dev.

Claude Cold Email: Prompts, a Reusable Skill, and Getting It Delivered

Ahmed Boateng — Sun, 18 Oct 2026 18:49:37 +0000

Liquid syntax error: Unknown tag 'endraw'

Claude Code for Sales: Build a Prospecting Agent That Finds, Verifies and Drafts in One Run

Ahmed Boateng — Sun, 18 Oct 2026 18:44:15 +0000

Liquid syntax error: Unknown tag 'endraw'

How to Use Claude for Sales Prospecting: 6 Workflows With the Exact Prompts

Ahmed Boateng — Sun, 18 Oct 2026 18:39:54 +0000

The short answer. Connect Claude to a go-to-market data source, then ask for the job in plain English with a credit budget. We ran six prospecting prompts: companies hiring sales right now, a one-page call brief, buyers complaining about a rival, every sales leader on a busy LinkedIn post, the sales lead at five named accounts, and which people on a list to contact first. All six came back with work you could act on the same day, for 336 credits in total.

How do you use Claude for sales prospecting?

To use Claude for sales prospecting, connect it to live sales data, then ask for the result you want in a normal sentence, with a budget on the end. Claude does the searching, the reading and the throwing-away. You get the shortlist.

We wanted to know how far that goes, so we gave six everyday prospecting jobs to a fresh Claude with no instructions beyond the prompt. It had one data source: Layrcake, which plugs into Claude as a connector through MCP (Model Context Protocol) and covers people and company search, work emails and verification, company research, and public LinkedIn, X and Reddit activity. Connecting it takes about two minutes, and the API reference lists every tool these prompts used.

Then we watched what it did. Some of it was what a good SDR does on a good day. Some of it was better, because it never gets bored of reading 563 LinkedIn headlines.

6 Jobs, one prompt each
336 Credits for all six ($3.36)
563 Headlines read in one go
0 Emails sent behind your back

1. Which companies are buying right now?

A company that's hiring salespeople has just admitted, in public, that it needs more pipeline. That's about as close to a buying signal as job boards get.

Claude

Find UK software companies with 50 to 500 staff that are showing buying signals right now: hiring for sales roles, or raised funding recently. Give me the 5 strongest, with the signal, its date and a one-line reason each is worth contacting this week. Budget: 35 credits.

Claude turned that into one search: UK, software, 50 to 500 staff, open sales roles posted in the last month. Four companies came back, and every one had posted a sales job in the previous two days. It ranked them by how hard they were hiring. The winner had posted 15 jobs in a month and was building a sales development team from scratch. Another was hiring for a region it hadn't sold into before.

Workflow 1 · companies hiring sales: 4 UK software companies building their sales teams this week

4 Companies · ≤ 2 days Since the job post · 15 Jobs in 30 days, top pick · 24 Credits

01: Workforce-scheduling software, 60 staff (SDR role · 15 posts in 30 days) · New team
02: Identity-verification software, 155 staff (Senior AE · new region) · New territory
03: Compliance software, 250 staff (Senior Client AE · 7 open roles) · Scaling
04: Contract software, 122 staff (Senior AE) · Hiring

Names and job-post links in the real answer · Charged per company returned

Swap the ask for "find me 5 companies like our best customer" and Claude matches on the same traits instead: country, size, a sales team that's growing. That version cost 30 credits.

2. Walk into a sales call already briefed

Everyone means to research before a call. Then the call before it overruns. This is the version that fits in the two minutes you have.

Claude

I have a call tomorrow with the head of sales at Pleo (pleo.io). Give me a one-page brief: what the company does and how it's doing, anything recent worth mentioning, who I'm likely speaking to and what they've been posting about, and three smart questions to ask. Budget: 35 credits.

The first thing Claude did was price its options. The deluxe one, a full picture of company and person, would have eaten 33 of the 35 credits. It took one look and went for the cheaper route: research the company, find the sales leaders, read what they'd posted.

The brief that came back read like it was written by someone who'd done their homework the night before. Funding to date, and a revenue figure clearly marked as an estimate. Two moves from the last four months, an AI launch and a new accounting partnership, each with a date and a source. The three people who run sales, by region, and a recent post from one of them about hiring. Then three questions that only make sense if you've read all of the above. Our favourite:

One of the three questions

With the June AI-agents launch, how has the buyer conversation shifted from card and expense tooling to the strategic finance story? Does your team sell it differently now?

Total: 11.5 credits.

3. Find buyers who are fed up with a competitor

People don't post when their software works. They post when it doesn't. And now and then, someone posts the whole shopping list.

Claude

We sell a sales engagement platform for B2B teams. Find recent posts on LinkedIn, X or Reddit from people frustrated with the sales engagement tool they use today (Outreach, Salesloft, Apollo or similar). Show me the 5 best, why each is a good lead, and how you'd open a conversation without being pushy. Budget: 20 credits.

Claude searched one rival at a time, and its checks threw out most of what it found: vendor marketing, tips, people being perfectly happy. What survived was a short list of the real thing. Top of it: a Reddit thread from the only salesperson at an 80-person company, openly shopping for a replacement, with their current stack, their budget worries and a wish list all spelled out, and nine comments already underway. Second: a LinkedIn post from someone ranking the tools in your category.

Claude's advice for the Reddit thread was the best line in the whole run: answer the question they asked, and leave the pitch at home.

Why this one is quieter. Public complaints that name a specific tool are rare, so expect a handful of strong leads a month. The ones you find are often mid-decision. The whole run cost 11 credits.

4. Turn one popular LinkedIn post into warm leads

If your buyers follow someone, the people reacting to that someone are your buyers, sitting in one place, and they've all just told you what they find interesting.

Claude

Gong's CEO Amit Bendov posts a lot about sales leadership, and heads of sales follow him. Look at his recent LinkedIn posts (his profile is https://www.linkedin.com/in/amitbendov/), pick the one with the most reactions and comments, and find everyone who engaged with it whose headline says head of sales, VP sales, sales director, CRO or VP revenue. Don't just sample it, I want all of them. Those are my warm leads. Budget: 280 credits.

Claude lined up the CEO's last seven posts and picked the clear winner: 563 people had reacted or commented. It took a quick sample first. That came back with a price for reading everyone (254 credits) and an estimate of three minutes. It set a 260-credit cap and sent the full read to run in the background.

Three minutes later: 563 people read, 26 sales leaders kept. Heads of sales, sales directors, VPs of sales, CROs. Six of them had titles that didn't use the exact words in the prompt (four were Chief Commercial Officers), and a title check kept them anyway, because a CCO is who you'd want. Each of those six is marked, so you can see exactly why it's there.

Workflow 4 · warm leads from one post

563 Engaged with the post → 548 People, after duplicates → 26 Sales leaders kept

Sales Directors: 7 (1 added by the title check)
Chief Revenue Officers: 5
VPs of Sales: 5 (1 added by the title check)
Chief Commercial and Sales Officers: 4 (4 added by the title check)
Other revenue leaders: 4
Heads of Sales: 1

Read in the background in about two and a half minutes, for 232.2 credits. Every person comes back with a profile link and what they did on the post.

Tool calls · the run (about 3 minutes)

find_posts · author: the CEO's profile · last 30 days → 7 posts · 4.5 cr
find_posts · engagers_of: top post · title filter · sample → 563 engaged · quote 254 cr · 4.1 cr
find_posts · same · read_all · budget 260 → background read
get_operation · until finished → 563 read · 26 kept · 232.2 cr

Twenty-six warm sales leaders, all with a profile link and a reason to talk to you, for 241 credits. That's the most expensive job in this guide, and probably the best value.

5. Find the right person at each account

You know the accounts. You don't know who to email or whether the address will bounce. Here are five companies, from an expense platform to an AI video startup.

Claude

Who runs sales at each of these companies? I want one person per company with a verified work email: pleo.io, syskit.com, jobadder.com, helpthemove.co.uk, d-id.com. Budget: 40 credits.

Claude found every company's sales leaders in a single search, then picked one per company. Where there was a choice, it said so: one company had a sales director, a head of sales and a regional head, and Claude told us which it chose and why, so we could swap. Five companies, five people from Head of Sales to CRO, five work emails, and all five came back valid. 31 credits.

Workflow 5 · the right person at each account: One sales leader per company, every email verified

5 / 5 Companies covered · 5 / 5 Emails valid · 1 People search · 31 Credits

[name removed]: Sales Director (expense-management platform) · Valid email
[name removed]: Head of Sales (Microsoft 365 governance software) · Valid email
[name removed]: Chief Revenue Officer (recruitment software) · Valid email
[name removed]: Chief Revenue Officer (home-moving services) · Valid email
[name removed]: Chief Revenue Officer (AI video platform) · Valid email

Names removed · 3 credits to search · 5 per email found

6. Decide who on your list to contact first

Everyone has a list of names. The trick is knowing who to call this week.

Claude

Here are 6 people on my list. Who's been active on LinkedIn lately, and what about? Rank them by who I should contact first this week, with a suggested first line for each. Budget: 20 credits.
- [Name], [Company], [LinkedIn profile URL]
- [Name], [Company], [LinkedIn profile URL]
- … (one line per person)

Claude checked what all six had been up to on LinkedIn. Every one of them had something. One had posted that morning, thanking their team for a record quarter: the easiest call of the week. Another had posted three times in five days. A third had only reacted to a webinar invite, which Claude rated medium: a small signal, but about exactly what they sell.

It ranked them on how fresh and how relevant each signal was, wrote a first line for each that refers to it, and read past anything personal to find the work post. 18 credits; anyone with nothing recent would have been free.

Workflow 6 · who to contact first: Six people, ranked by what they did on LinkedIn this month

01: Posted that morning: a record quarter, new customers (home-moving services) · First
02: Reacted to a webinar on their own product area (Microsoft 365 software) · This week
03: Three posts in five days on AI in recruitment (recruitment software) · This week
04: Posted on AI in business development (print-management software) · This week
05: Reacted to someone else's post (design-system software) · Later
06: Reacted to someone else's post (AI video platform) · Later

Names removed · Each comes with a first line that refers to it · 18 credits

What makes a prospecting prompt work?

Look back at the six prompts. Each one says four things, and the runs went best when all four were there. Want all six in one place? Download the prompt pack, with the parts to change in [brackets].

The result you want. "The 5 strongest." "A one-page brief." "One person per company." Claude plans backwards from it.

The filters you'd use yourself. Country, size, titles, a competitor's name. Each word becomes a search filter, so vague in means vague out.

What you want next to each result. "With the signal and its date." "Why each is a good lead." "A first line." That's the difference between a list and a to-do.

A budget. Claude checks the cost of every call and stops before it runs out. Searches that find nothing are free, so being ambitious costs little.

What does it cost?

All six jobs together came to 336 credits, about $3.36 at $0.01 per credit (at the prices on the days we ran them, 2–3 October 2026). The full read of 563 people was most of that. The other five cost between 11 and 31 credits each, so somewhere between a coffee and half a coffee.

What	Price
Companies hiring, or like a customer	11 credits per hiring company, 13 per lookalike
Company research	9 credits
People search	0.25 credits per person (3 minimum)
Work email	5 credits when found, free when not
Recent activity for a list	3 credits per person with activity
People who engaged with a post	0.5 credits per person kept, plus 0.4 per person read with a title filter

Nothing sends on its own. These prompts research and draft. Creating a campaign is a separate step, and launching it asks a signed-in person to approve before a single email leaves.

FAQ

Can Claude do sales prospecting?

Yes, once it's connected to live data. With a go-to-market data connector, Claude can find companies and people, look up and verify work emails, research a company and read public LinkedIn, X and Reddit activity. In our tests six plain-English prompts each came back with results you could act on the same day.

What is a good Claude prompt for prospecting?

Say the outcome you want, the filters you would use yourself, what you want next to each result, and a credit budget. For example: find UK software companies with 50 to 500 staff hiring for sales roles, give me the 5 strongest with the signal and its date, budget 60 credits.

How much does prospecting with Claude cost?

In our runs, five of the six workflows cost between 11 and 31 credits (11 to 31 cents). Reading everyone who engaged with a 563-person LinkedIn post cost 241 credits. Searches that find nothing are free.

Does Claude send emails to prospects by itself?

No. These prompts only research and draft. Creating a campaign is a separate step, and launching it asks a signed-in person to approve before anything sends.

Download: Prompt pack

Originally published at layrcake.dev.

How to Use Claude for Lead Generation: The Full Loop, Tested (2026)

Ahmed Boateng — Sun, 18 Oct 2026 18:39:17 +0000

The short answer. Claude can run lead generation end to end once it has tools for finding people, checking their emails and drafting a sequence. Connect a data layer over MCP, describe who you want in plain English, verify before you draft, and approve the campaign yourself before anything sends. Our run took 20 leads to 15 ready-to-send drafts for 145 credits.

What we ran

We tested Claude for lead generation end to end. The job: find 20 heads of sales at London SaaS companies, check their emails, find a reason to write to each of them, and draft a sequence we would be happy to send. Claude, running in Claude Code, made every call itself against live data on 1 October 2026. We logged each call and what it cost.

20 Leads asked for
15 Verified and drafted
15 of 15 With a recent reason to write
145 Credits used ($1.45)

The job we gave Claude

Find 20 heads of sales at SaaS companies in London with 20 to 200 staff. Verify every email, find a recent reason to reach out to each one, and draft a three-step sequence that opens with it. Stop before sending.

What does Claude need to do lead generation?

Claude, Anthropic's AI assistant, needs live data and a way to act on it. On its own it can write a good email and reason about who to target, but it cannot look up who works where today or whether an address will bounce.

An MCP (Model Context Protocol) server gives it those tools. We used Layrcake, which puts company search, people search, email verification, social posts and campaign drafting behind one connection, so Claude could run the loop without switching tools.

Step 1: Connect the data layer

In Claude Code it is one command and a sign-in. Run the command, then type /mcp, choose layrcake and choose Authenticate. Your browser opens to sign in and approve access. There is no key to paste.

Terminal

claude mcp add --transport http --scope user layrcake https://mcp.layrcake.dev/mcp

In the Claude app on the web or desktop, open Settings, then Connectors, choose Add custom connector and paste https://mcp.layrcake.dev/mcp. The Claude Code setup page walks through the first run, and the connect page has the steps for other apps.

Step 2: Describe who you want

Claude turned the sentence into one people search: the title, the city, the industry and the company size, with a limit of 20. The search matched 91 people and returned the first 20 in 1.8 seconds, for 4 credits.

find_leads arguments · exactly as run

{
  "audience": "Heads of sales at SaaS companies in London with 20 to 200 staff",
  "titles": ["Head of Sales"],
  "locations": ["London, United Kingdom"],
  "industries": ["Software Development"],
  "company_headcount": { "min": 20, "max": 200 },
  "limit": 20,
  "budget_credits": 10
}

Search results carry a name, a title, a company and a LinkedIn profile. They do not carry email addresses, so Claude's next call was enrich, once per person, asking only for a work email. Enrich found 16 of the 20. The four it could not find cost nothing; each hit cost 5 credits.

Three of the 20 were heads of sales development, who run the SDR team. We kept them for this run. If you only want the person who owns the number, that is the filter to tighten.

Tool calls · what Claude did (about 90 s of tool time)

find_leads · Head of Sales · London · Software Development · 20 to 200 staff · limit 20 → 20 of 91 matches · 4 cr
enrich × 20 · want: work_email → 16 found · 4 misses free · 80 cr
verify_email × 16 · one call per address → 10 valid · 5 catch-all · 1 invalid · 16 cr
find_posts · leads: 15 · latest activity per person → 15 with a signal · 45 cr
create_campaign · 3 steps · 15 leads → draft · 0 cr

Every tool's arguments and result fields are in the API reference.

Step 3: Verify before you draft

We asked Claude to verify every address before writing a word. Our view: a bounced first email spends sending reputation you need for the rest of the campaign, so the check comes first.

Enrich had labelled all 16 addresses valid. The second check disagreed on six of them. One address was invalid. Five sat on catch-all domains, which accept mail for any address, so a check can't confirm the mailbox exists. Claude dropped the invalid address and kept the catch-all ones, a judgement call we were comfortable with for a first send. That left 15.

One more thing the run turned up: one address enrich found was at a different company from the one the search listed. Search data can lag a job move, which is why we check the address we are about to use.

Step	Asked	Hits	Credits
Find leads	20	20	4
Find work emails	20	16	80
Verify emails	16	15 kept	16
Recent activity	15	15	45
Draft campaign	15	15	0

Searches and lookups that found nothing were free. Every verification cost 1 credit, the failed one included. Every call reported its charge in the response.

Step 4: Draft the sequence and stop at approval

Before writing, Claude looked for a recent reason to contact each person. One find_posts call took the whole list of 15 and returned, for each person, the best recent activity it could find: their own post, a comment, a reaction to someone else's post, or failing those, a post from their company. Lists over 10 run in the background, so Claude polled for the result. It took 38 seconds.

All 15 came back with something: 9 of their own posts, 5 reactions and 1 company post. The newest was from that morning and the oldest was 23 days old, with a median of 7 days. Each lead with a signal cost 3 credits. A lead with nothing would have cost nothing.

Claude wrote one opening line per person from that activity and passed it to the campaign as a custom field, so the first email opens with it. Own posts made the strongest openers. Reactions are weaker, because liking a post says less than writing one, so those lines say "saw you liked" and stop there.

Email 1 · template as drafted ([fields] fill per lead; the product is a placeholder for the test)

Subject: Sales calls at [company_name]

Hi [first_name],

[opener]

Most heads of sales I speak to can say how many calls the team made last week, but not which ones moved a deal forward.

We built a call-recording tool that pulls the moments that matter out of every call (objections, pricing questions, next steps) and puts them in one weekly summary for you.

Would a look at last week's summary for a team your size be useful?

Thanks,
Sam

Opening line · synthetic example

Saw your post about hiring two more AEs for the EMEA team. Sounds like a big quarter.

Then create_campaign built the draft: three emails over seven days and 15 leads accepted, with none removed by the opt-out and suppression checks. It cost 0 credits. Creating a campaign never sends anything. Sending takes a second call, launch_campaign, which answers with a preview, a fresh check of the recipients and an approval link.

Nothing sends on its own. launch_campaign returns a preview and an approval link. A signed-in person approves before a single email leaves. We stopped at the draft, and so does this guide.

What did it cost?

The whole run cost 145 credits, about $1.45 at $0.01 per credit (at the prices on the day we ran it, 1 October 2026). That is a little under 10 credits for each of the 15 people who ended up in the draft with a checked address and a reason to write.

Finding emails was most of the bill: 80 of the 145 credits. The tool calls took about 90 seconds between them. The run took under ten minutes end to end, most of it spent writing and reading the openers.

Before you hit send

Claude did the legwork. Three things stay with you, and none takes long.

Read the openers. They come from real posts, so they are specific. Tweak any that don't sound like you.

Decide on catch-all addresses. They accept mail for any name, so a check can't confirm the mailbox. Keep them, or drop them if your sending domain is new.

Approve the launch. Layrcake does not launch a campaign until a signed-in person approves it. For anything that emails real people, that is the default we want.

To rerun this every week, turn the loop into a Claude Code skill.

FAQ

Can Claude generate leads on its own?

Not from its training data. Claude needs tools that search live company and people data and check email addresses. Connected over MCP, it calls those tools and works with what they return.

Is it safe to let Claude send cold email?

Only with an approval step. In our run the campaign was created as a draft. Launching it returns a preview and an approval link, and nothing sends until a signed-in person approves.

What does a run like this cost?

Our run cost 145 credits (about $1.45) for 20 leads, which left 15 ready to send. Searches and email lookups that find nothing are free; an email check that returns an answer costs 1.25 credits.

Does this work in the Claude app as well as Claude Code?

Yes. Claude on the web and desktop connects to the same MCP server as a custom connector; Claude Code adds it with one command.

Originally published at layrcake.dev.