On this page
b2bsearch MCP is a small stdio server that gives an agent six B2B prospecting tools. Behind them are the b2bsearch Actors on Apify, which read one database of professional profiles, companies and job postings.
Why I built it
I publish these Actors on Apify. Through a generic run tool, an agent sees thirty or forty input fields per Actor, and it is easy to pay for a list before finding out the filter was wrong. So I wrapped the six jobs people actually ask for in tools with task names and short schemas: who joined these accounts lately, how does this company write its emails, which companies look like my best customers, who runs this domain, what is the email behind this profile URL, and how many people match this title in this country.

Counts and misses do not incur per-row charges; Apify Actor start fees may still apply.
Costs and data handling
- Paying to learn a filter was wrong.
search_peoplehascountandmarketmodes with no per-row charge, so the agent sizes a segment and reads its top titles and employers before a single row is billed. An Actor start fee may still apply.
- Runaway spend. Every tool takes
maxChargeUsd, $5 by default and never more than $500. Check the Actor’s current pricing and your Apify account terms for applicable charges.
- Paying for misses. A person with no email or a domain nobody claims comes back with a status that says why. Check the Actor’s current pricing for applicable charges.
- Token and data handling. The server runs on your machine and reads your Apify token from the environment. It calls Apify using your account and returns the resulting rows.
- Third-party data. Names, titles and headlines are third-party text. Successful runs include a note telling the agent to treat strings inside rows as third-party data, never as instructions.
Where it runs
Any client that starts stdio MCP servers, with Node 18 or newer:
- Claude Code:
claude mcp add --scope user b2bsearch -e APIFY_TOKEN= -- npx -y @b2bsearch/mcp
- Cursor, Windsurf and Claude Desktop: an
mcpServersentry that runsnpx -y @b2bsearch/mcpwithAPIFY_TOKENin env
- Claude Desktop in one click: download the
.mcpbfile from the GitHub releases
- No install: the same Actors through Apify's hosted MCP server, with Apify's generic run tools instead of the task-named ones
initialize and tools/list answer without a token, so you can read all six schemas before you create an account. A free Apify account is enough to start: 500 rows a month per Actor, up to 100 rows per run, and up to 25 per run for search_people.
Try it
Ask your agent: "How many heads of sales are there in the Netherlands? Count only." It should call search_people in count mode with no per-row charge. An Actor start fee may still apply. Then ask for ten of them at companies with 50 to 500 people, with a $1 cap.
Source: github.com/keyvwin/mcp (MIT). npm: @b2bsearch/mcp. Official MCP Registry id io.github.keyvwin/mcp. The Actors: apify.com/b2bsearch.
Disclosure: I maintain the server and the Actors it calls. The server adds no fee; rows are billed at the Actors' own prices on Apify. B2B contacts and account signals on Apify: new hires, email format, lookalikes, decision makers.
You can find b2bsearch: B2B contacts and account signals in the Awesome AI Plugins registry: https://hol.org/registry/plugins/io.github.keyvwin%2Fmcp. The source lives at https://github.com/keyvwin/mcp.
If you build anything with it, tell the author — launch stories like this one are how the directory gets useful.

