Crustdata

Search and enrich B2B company and people data.

Opens Agent Studio, where connecting is one click. The connector URL below works in any MCP client.

Paste it into any MCP client. Setup docs

What the Crustdata connector does

The Crustdata MCP server connects a B2B company and people database to Claude, ChatGPT and any MCP-compatible agent. Anthropic's directory lists 23 tools that search and enrich companies, individual people, jobs and social posts. Sign-in is OAuth. Nearly every tool spends credits, and the people tools return named individuals.

Verified connector

Listed by Anthropic as a partner connector in its Connectors Directory.

Connection checked by Agentman on .

Anthropic states this reflects the level of review a connector received, not a security audit.

Crustdata tools (23)

  • crustdata_company_identify
  • crustdata_company_enrich
  • crustdata_company_search_db
  • crustdata_company_search
  • crustdata_company_linkedin_posts
  • crustdata_job_search
  • crustdata_job_search_live
  • crustdata_batch_job_search
  • crustdata_batch_job_search_live
  • crustdata_people_enrich
  • crustdata_people_search_db
  • crustdata_people_search
  • crustdata_linkedin_posts
  • crustdata_credits_check
  • crustdata_autocomplete_company
  • crustdata_autocomplete_person
  • crustdata_autocomplete_linkedin_filter
  • crustdata_watcher_create
  • crustdata_watcher_update
  • crustdata_watcher_simulate
  • crustdata_watcher_list
  • crustdata_web_search
  • crustdata_web_fetch

Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.

Limits

  • The scope boundary sits in the right place, but nothing enforces which tools you granted. The server's RFC 9728 descriptor declares scopes_supported as read and write, verified live on 2026-08-23 — a genuine two-value split, unlike the empty arrays and identity-only claims most connectors ship. It separates the 21 reads from the two watcher writes. It does not separate a company lookup from a personal-phone-number lookup, which is the boundary a privacy-conscious reader would actually want. We did not verify that the server enforces the scope it advertises.
  • Every people tool is a personal-data tool. There is no aggregate-only mode. If the risk you are managing is exposure of individuals' contact details to an agent, this connector cannot be configured to avoid it.
  • Crustdata names none of its data sources. "15+ sources" is the entire published claim.
  • No public data licence governs connector output. The public Terms page covers the website only. Your commercial agreement is the operative document and we have not seen it.
  • A watcher outlives the conversation. crustdata_watcher_create starts a recurring job with no documented expiry by default, and no tool in the listing deletes one.
  • Costs are per result, not per call. A broad search bills for every record it returns, and the agent does not see the bill until after the call.
  • Rate limits are per endpoint and low on the expensive ones. Crustdata documents 15 requests per minute on person enrich, 30 on person search, and 5 to 10 on the live and batch endpoints, returning HTTP 429 when exceeded. Limits vary by plan.
  • Some fields need account-level permission. Crustdata's person enrich reference marks certifications, honors and the permanent URN-based profile URL as requiring field-level permission on the API key, so two accounts can get different results from the same call.
  • We could not read tool schemas or safety annotations. The endpoint returned 401 to an anonymous request, so no parameter-level detail on this page comes from the server itself.
  • A second Crustdata MCP server exists and is a dispatcher. The vendor's own install links point at it. Nothing on this page describes its reachable surface.
  • The support address differs between sources. Anthropic's directory and Crustdata's older MCP page give support@crustdata.com; its newer pages give support@crustdata.co. Both appear in first-party material.

Frequently asked questions

Does the Crustdata MCP server return data about individual named people?

Yes, directly and by design. Crustdata's own documentation describes 800 million professional profiles, and the person enrich response returns a basic_profile section carrying name, first name, last name, headline, current title, location and profile picture, plus full employment history and education. This is not an aggregate surface with a person-level dimension hidden inside it. Named individuals are the product.

Can the Crustdata connector return someone's personal email address or phone number?

Yes. Crustdata's contact enrichment documents three requestable contact types: business emails at 1 credit per matched person, personal emails at 2 credits, and phone numbers at 2 credits. The connector's people enrich tool reaches this surface. Crustdata's own MCP page lists the example prompt asking Claude to get the email for a person's profile.

Where does Crustdata's people data come from?

Crustdata does not publish a source list. Its people dataset page claims 1 billion people from 15-plus sources with 95-plus datapoints per person, but never names those sources. The API is organised around professional network profile URLs as the primary identifier, and the docs describe real-time retrieval from the web. We could not establish whether any given record was scraped, licensed or contributed.

Can I redistribute what the Crustdata connector returns?

Crustdata publishes no data licence answering this. Its only public Terms page is website-materials boilerplate governing downloads from crustdata.com, granting a personal, non-commercial viewing licence to the website and saying nothing about API or connector output. The contract that actually governs your data use is your commercial agreement, which is not public. Ask before publishing connector output.

Does one Crustdata tool call spend money?

Most of them do. Crustdata prices per result and per record, not per call: person search at 0.03 credits per result, person enrich at 1 to 7 per record, company enrich at 2 to 4, web search and fetch at 1 each, and a person entity watcher at 5 to 150 credits per changed person. Identify and autocomplete are free.

Can the agent check its Crustdata credit balance before spending?

Yes. The crustdata_credits_check tool is free and reads the account balance, the recurring grant and the next refresh date. Crustdata also returns an X-Credits-Used header on every API request. This is the accounting tool that many metered connectors omit, and it means an agent can be instructed to check the balance before a large batch.

Does the Crustdata connector write, delete or send anything?

It creates and updates watchers, and nothing else. Three of the 23 tools mutate state: crustdata_watcher_create, crustdata_watcher_update and, on the vendor's own reading, nothing destructive. No tool deletes a watcher. The RFC 9728 descriptor advertises a write scope, verified live on 2026-08-23, which matches these two names rather than pointing at a hidden surface.

What does creating a Crustdata watcher actually do?

It starts recurring surveillance of named people and pushes it to a URL you choose. Crustdata documents watcher notification channels as a webhook with an arbitrary url and arbitrary headers, a Slack webhook, or Google Chat. A person entity watcher re-runs on a schedule, compares each watched person against their last snapshot, and delivers the changed profile fields outward.

Sources

Use it in an agent

Put Crustdata to work.

Connect Crustdata once and your agents call these tools on their own: on a schedule, in a workflow, with nobody at the keyboard.

Call (650) 285-1019Our AI receptionist answers.