Summary

Metered data done honestly, including the advice that costs them a call.

Two things are unusually well handled. Costs are visible in the response rather than on an invoice later, and a discrepancy between ids sent and records billed tells you exactly how many were not found. And the documentation actively steers you away from the expensive habit: reading a large file back page by page is free in credits but re-billed as input tokens every turn, so it tells you to get downloads working and keep the paging tool for small reads. It also says, in bold, never to re-run a fetch to recover a file you already paid for.

What it is

A remote MCP server over Coresignal's B2B data, connected with OAuth 2.1 through your Coresignal dashboard account so no API key ever goes into a config file. Searches are written in plain language, every result row shows why it matched, and large result sets are delivered as files rather than dumped into the conversation.

What you get

`entity_search` — natural-language search over employee, company or job records, returning the exact `total_count` of matches, up to 20 preview rows each showing the fields the query matched on, and a `cache_id` handle valid for an hour that lets a fetch collect the matched records without re-running or re-paying for the search
`entity_fields` — free keyword lookup over an entity's roughly 300 field names, so you can build a custom fetch scope without loading the whole vocabulary into context
`entity_fetch` — full or projected records, either up to 20 hand-picked ids or up to 1,000 records per call via a cache handle
`email_enrich` — verified, deliverable business emails for up to 1,000 employee ids; EEA and UK contacts are not accessible under GDPR
`artifact_read` — pages a delivered file back into the conversation, free in credits, which is what makes file delivery work in clients that cannot download or touch a filesystem
Cost transparency as a design property: every response reports `credits_consumed`, expensive calls confirm before spending, fetches confirm field scope first, and a fetch that cannot be delivered fails before any credits are spent
Published pricing per action — 20 credits per search at any limit including zero, 20 per employee or company record found, 1 per job record, 10 per email actually found with misses free, and field lookups and artifact reads free

Requirements

A Coresignal account with an active subscription and a team API key provisioned in the dashboard — the server resolves your team's key after sign-in, so you never paste it anywhere. There is a 7-day free trial including 2,000 credits. Connect any MCP client to `https://mcp.coresignal.com/mcp/v2`; the first connection opens a browser to sign in, and that is the entire setup — no environment variables, no secrets. Claude Code adds it with `claude mcp add --transport http coresignal https://mcp.coresignal.com/mcp/v2`, and Claude Desktop, Codex, Cursor, OpenCode, VS Code and Cline each have a documented path. **One thing to configure up front in Claude Desktop**: add `mcp.coresignal.com` to the domain allowlist, or download links from the server are blocked — you can still read data back in-chat with the free artifact tool, but downloads are the cheaper path.

Setup effort

One command plus a key — claude mcp add --transport http coresignal https://mcp.coresignal.com/mcp/v2, then supply credentials