Developers & AI agents

The PeopleSearch.im API & MCP server

The same people-search backend as the app, for your code and your AI agents. Call it over plain HTTP, or connect it as a Model Context Protocol server so Claude, ChatGPT, Cursor, or any agent can search people, verify emails, and fetch verified work emails, billed to the same credits: searching is free, 1 credit per profile.

MCP server REST + JSONOAuth or API keySearch without a key1 credit / profile

New accounts get 10 free credits. Searching is always free; you only pay for the profiles and emails you keep.

Try it now, no key needed

curl -X POST https://peoplesearch.im/api/v1/search \
  -H "Content-Type: application/json" \
  -d '{"query":"Heads of Marketing at SaaS companies in New York"}'

Or open it as a plain URL: https://peoplesearch.im/api/v1/search?q=Heads+of+Marketing+at+SaaS+companies+in+New+York(opens in a new tab)

You get the true total and up to 3 masked previews, up to 10 searches a day. A free key returns full pages and unlocks names, LinkedIn profiles and verified work emails.

Use with AI agents

Connect it to Claude, ChatGPT, Cursor, or any MCP client

PeopleSearch.im is a remote Model Context Protocol server. Add its URL to your AI app and approve the connection when it asks: there is no key to copy. Your assistant can then search people and companies, reveal profiles, and fetch verified work emails on your behalf, spending your credits (the same balance as the API and the app).

https://peoplesearch.im/api/mcp

New to it? The MCP server overview covers what an agent can do with it and the use cases. For the bigger picture, our guide to giving AI agents access to people data covers how MCP and REST fit together, the tradeoffs, and keeping an agent's data grounded and compliant.

Claude (claude.ai, Claude Desktop, and mobile)

Open Settings, then Connectors, and choose Add custom connector. Name it PeopleSearch.im, paste the URL above, and add it. Searching works straight away, a few times a day. When Claude needs your account (to reveal a profile or fetch an email, or once the free searches are used) it shows a Connect button: sign in, approve, and the request continues.

ChatGPT

PeopleSearch.im is in ChatGPT's plugin directory: open PeopleSearch.im in ChatGPT and select Install plugin. Then ask for the people you need in any chat, and approve the connection when it asks you to sign in. Your ChatGPT workspace settings decide which plugins can be added.

Claude Code

claude mcp add --transport http peoplesearch https://peoplesearch.im/api/mcp

Then run /mcp inside Claude Code and authenticate PeopleSearch.im to connect your account.

Cursor (~/.cursor/mcp.json)

{
  "mcpServers": {
    "peoplesearch": {
      "url": "https://peoplesearch.im/api/mcp"
    }
  }
}

VS Code (.vscode/mcp.json)

{
  "servers": {
    "peoplesearch": {
      "type": "http",
      "url": "https://peoplesearch.im/api/mcp"
    }
  }
}

Prefer an API key? (Windsurf, scripts, and clients without OAuth)

Clients that support MCP authorization ask you to sign in and approve when a tool needs your account. If yours does not, create a free account to get a key and send it as a header instead:

claude mcp add --transport http peoplesearch https://peoplesearch.im/api/mcp \
  --header "Authorization: Bearer sk_sift_..."
{
  "mcpServers": {
    "peoplesearch": {
      "serverUrl": "https://peoplesearch.im/api/mcp",
      "headers": { "Authorization": "Bearer sk_sift_..." }
    }
  }
}

For a stdio-only client, bridge it with mcp-remote:

{
  "mcpServers": {
    "peoplesearch": {
      "command": "npx",
      "args": [
        "-y", "mcp-remote", "https://peoplesearch.im/api/mcp",
        "--header", "Authorization: Bearer sk_sift_..."
      ]
    }
  }
}

The tools your agent gets

  • people_searchSearch peoplefree
  • reveal_profileReveal a person1 credit
  • find_peopleFind & unlock a batch of people1 credit
  • fetch_emailFetch a verified work email1 credit
  • find_work_emailFind a person's work email2 credits
  • verify_emailVerify an email address1 credit
  • find_company_emailsFind a company's emails2 credits
  • reverse_email_lookupReverse email lookup13 credits
  • find_linkedin_profileFind a LinkedIn profile2 credits
  • lookup_linkedin_profileLook up a LinkedIn profile2 credits
  • find_email_from_linkedinFind the email behind a LinkedIn profile3 credits
  • lookup_companyLook up a company1 credit
  • company_searchSearch companiesfree
  • check_creditsCheck credit balancefree

Free tools spend no credits. Paid tools charge on a billable result and refund automatically when nothing is found. Machine-readable spec: /openapi.json.

Authentication

Use a Bearer key for the REST API

Create a free account to get a key, then send it on every request:

Authorization: Bearer sk_sift_...

An X-API-Key: sk_sift_... header works too. The API answers browsers on any origin, so you can try it straight from front-end code, but a key in a web page is readable by anyone who loads it and spends your credits. For anything public, call the API from your backend, and give each app its own key so you can revoke one without the others.

Search is the one call that also answers with no key, so you can try the API first: up to 3 results per search and 10 searches a day per IP address, within a daily limit that all callers without a key share, with a keyless field in the response saying so. Past any limit it returns 429. A tool that searches for many people should send a key: without one, each app (named by its User-Agent) gets only a share of that daily limit, from however many addresses it calls. Every other endpoint needs a key.

A 503 with provider_unavailable means people data is temporarily unavailable on our side: nothing was charged, your request was fine, and the Retry-After header says when to ask again.

Endpoint

Search

POST /api/v1/searchGET /api/v1/search?q=

Describe who you want in plain English, or pass exact filters (job titles, companies, LinkedIn company pages, domains, locations, industries, schools, skills, languages, company size, years of experience). Returns a free sample of masked previews and the true total, so you can validate the match before you pay, then deliver from the same query or filters (below). Searching is free.

curl -X POST https://peoplesearch.im/api/v1/search \
  -H "Authorization: Bearer sk_sift_..." \
  -H "Content-Type: application/json" \
  -d '{"query":"Heads of Marketing at Series B SaaS in New York","limit":10}'

The same search also works as a plain URL, for a browser or any tool that can only open links: GET /api/v1/search with the words in q (and optional limit and page). It needs no key either; a key, when you send one, goes in the header, never in the URL.

curl "https://peoplesearch.im/api/v1/search?q=Heads+of+Marketing+at+SaaS+companies+in+New+York"

Filters are matched as given, nothing is re-interpreted, so programs and AI agents should prefer them. Values in one list are alternatives, different fields must all match. Titles, companies and schools match as substrings; places are matched in every spelling profiles use (Germany also finds Deutschland); industries use LinkedIn's names. Anything that could not be applied comes back in ignored. The full schema is PeopleFilters in openapi.json.

curl -X POST https://peoplesearch.im/api/v1/search \
  -H "Authorization: Bearer sk_sift_..." \
  -H "Content-Type: application/json" \
  -d '{"filters":{"job_titles":["M&A Associate","Investment Banking Associate"],
               "locations":["Frankfurt","Munich"],
               "industries":["Investment Banking"],
               "company_size_min":50}}'
{
  "ok": true,
  "total": 4213,
  "results": [
    {
      "key": "q3xV9mL2Tf0bR7aKdW1sEu",
      "token": "sealed-token-you-pass-to-reveal",
      "maskedName": "Jordan M.",
      "title": "Head of Marketing",
      "industry": "Software Development",
      "companySize": 180,
      "location": "New York, United States"
    }
  ],
  "token": "next-page-token",
  "chips": [{ "label": "Role", "value": "Head of Marketing" }],
  "parser": "ai",
  "balance": 120
}

The same endpoint searches companies with "entity": "company": by name, website, LinkedIn page, industry, headquarters, size, type or keywords, in plain words or as CompanyFilters. Companies come back whole in companies (website, LinkedIn page, industry, employees, headquarters, description), free, up to 100 a page with a key.

curl -X POST https://peoplesearch.im/api/v1/search \
  -H "Authorization: Bearer sk_sift_..." \
  -H "Content-Type: application/json" \
  -d '{"entity":"company","limit":100,
       "filters":{"industries":["Software Development"],
                  "locations":["Berlin"],"size_min":50,"size_max":200}}'
Endpoint

Deliver people with their emails (1 credit per profile, 1 per email)

POST /api/v1/deliver

The paid step. Pass the same query or filters and how many profiles you want in count (default 100, up to 100 per call). Delivers the full profiles in ONE atomic charge: 1 credit each, and any you already own are free. Add "withEmail": true to look up each person's verified work email in the same call: 1 more credit per person, refunded when none is found. The call waits about 12 seconds for the emails; one still searching is collected free with the email endpoint below and its id, and emailsUnpaid counts any the balance could not cover. Returns 402 when the balance cannot cover the profiles.

curl -X POST https://peoplesearch.im/api/v1/deliver \
  -H "Authorization: Bearer sk_sift_..." \
  -H "Content-Type: application/json" \
  -d '{"query":"Heads of Marketing at Series B SaaS in New York","count":5,"withEmail":true}'
{
  "ok": true,
  "charged": 5,
  "balance": 1,
  "total": 4213,
  "emailsUnpaid": 0,
  "profiles": [
    {
      "id": "…",
      "profile": {
        "name": "Jordan Miller",
        "title": "Head of Marketing",
        "company": "Acme",
        "location": "New York, United States",
        "linkedinUrl": "https://linkedin.com/in/...",
        "email": "jordan.miller@acme.com",
        "emailStatus": "found",
        "emailCertainty": "ultra_sure"
      }
    }
  ]
}

Leave withEmail out to get the profiles alone (emails idle) and fetch only the emails you need later.

Endpoint

Reveal a single profile (1 credit)

POST /api/v1/reveal

For one profile you already have a search token for (prefer Deliver for volume). Spends one credit and returns the full profile (without the email). Add "withEmail": true to also fetch the professional email in the same call (a second credit, refunded if none is found); the endpoint waits briefly and, if the lookup is still running, returns emailStatus:"searching" so you can re-call the same token for free to get it once resolved. Revealing the same person again is free (idempotent). Returns 402 when out of credits.

curl -X POST https://peoplesearch.im/api/v1/reveal \
  -H "Authorization: Bearer sk_sift_..." \
  -H "Content-Type: application/json" \
  -d '{"token":"<token from a search result>"}'
{
  "ok": true,
  "id": "…",
  "balance": 119,
  "profile": {
    "name": "Jordan Miller",
    "title": "Head of Marketing",
    "company": "Acme",
    "companyIndustry": "Software Development",
    "companySize": 180,
    "location": "New York, United States",
    "linkedinUrl": "https://linkedin.com/in/...",
    "email": null,
    "emailStatus": "idle"
  }
}

With withEmail (a second credit, refunded if none is found):

curl -X POST https://peoplesearch.im/api/v1/reveal \
  -H "Authorization: Bearer sk_sift_..." \
  -H "Content-Type: application/json" \
  -d '{"token":"<token>","withEmail":true}'
# profile.email: "jordan@acme.com", emailStatus: "found", emailCertainty: "very_sure"
Endpoint

Email (1 credit, refunded if none)

POST /api/v1/email

Fetch a delivered (or revealed) profile's professional email by its id. Spends one credit, refunded if no email is found. Waits briefly; if the lookup is still running it returns emailStatus:"searching" so you can re-call to poll.

An address that accepts mail but carries nothing of the person's name (a first-name address such as laura@, which can belong to anyone of that first name there) comes back as emailStatus:"unconfirmed" with a note. It is not charged and is not counted as their email; a reverse email lookup says whose it is.

curl -X POST https://peoplesearch.im/api/v1/email \
  -H "Authorization: Bearer sk_sift_..." \
  -H "Content-Type: application/json" \
  -d '{"id":"<id from a deliver or reveal response>"}'
# { "ok": true, "email": "jordan@acme.com", "emailStatus": "found", "certainty": "very_sure", "balance": 19 }
Endpoint

Credit balance

GET /api/v1/credits

curl https://peoplesearch.im/api/v1/credits \
  -H "Authorization: Bearer sk_sift_..."
# { "ok": true, "balance": 119, "creditPerReveal": 1 }
No code

Use it in Zapier

Four steps for any Zap, on the same credits as the API

Find a person's work email from their name and company when a row lands in a sheet, verify an address before it enters your CRM, or pull people by job title, company, industry and location into your pipeline. The integration is in beta: open the invite, then choose PeopleSearch.im in a Zap and connect it with an API key from this page.

Find Work EmailVerify Email AddressFind PeopleGet Work Email
No code

Use it in Make

Seven modules for any scenario, plus a call to any endpoint, on the same credits as the API

Search people and companies for free inside a scenario, unlock the profiles you want, find and verify work emails, then send them on to a sheet or your CRM. Searches page through as many results as you ask for. Open the invite to install the app in your Make organization, then add a PeopleSearch.im module and connect it with an API key from this page.

Search peopleFind peopleGet a profileGet a work emailFind a work email by nameVerify an email addressSearch companies
Marketplace

Run it on Apify

The same people data runs as Actors in Apify Store, for teams that already run their data jobs on Apify or call Actors from an AI agent through Apify's MCP server. You pay Apify per result (a verified email only when one is found), with no PeopleSearch.im account or key.

From our own server

The MCP clients that connect, measured

In the last 30 days 497 different client programs connected to this MCP server. Each counts once, however many requests it made, so a directory that checks the server every minute weighs no more than a developer trying it once. 16% of them already asked for the newest revision, 2026-07-28.

497
Client programs connected
last 30 days
48%
Listed the tools
the rest only checked the server
4%
Called a tool
a search, a lookup
 Share of clientsBarClients
Protocol 2026-07-2816%78*
Protocol 2025-11-2531%153
Protocol 2025-06-1860%300
Protocol 2025-03-266%28*
Protocol 2024-11-054%20*
A client that asked for several revisions counts in each. Updated daily, last on 18 November 2026. Every figure: figures.csv.

Same credits as the app

New accounts get 10 free credits. After that, buy credits once and use them from the dashboard or the API. Top up any time.