Use FindLeadEmails in Cursor, Claude and other AI agents

Sign in and open Developers. Create a key, copy it when it is shown, and send it on every request as Authorization: Bearer YOUR_API_KEY. Only a hash is stored, so a lost key cannot be displayed again — rotate it on that page to get a new value.

Project config goes in .cursor/mcp.json. A file at ~/.cursor/mcp.json applies to every project. Cursor's remote-server entry is a url plus headers.

{
  "mcpServers": {
    "findleademails": {
      "url": "https://www.findleademails.com/api/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

To keep the key out of the file, Cursor will substitute ${env:FINDLEADEMAILS_API_KEY} inside headers. The copy-paste block above uses the placeholder so you can see the header shape.

Add to Cursor

That button is Cursor's install link: the server name is the name parameter, and config is the base64 of the same url and headers object. Cursor asks before it installs. The installed entry still contains YOUR_API_KEY — replace it with the key from Developers.

claude_desktop_config.json runs a local command. A remote Streamable HTTP server needs the mcp-remote bridge via npx (Node.js 18 or newer). The header is passed without a space after the colon because some clients split arguments on spaces; the bearer value, which contains a space, lives in env.

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json. Windows: %APPDATA%\Claude\claude_desktop_config.json. If the file is missing, enable it under Settings → Developer. Quit and reopen Claude Desktop after saving.

{
  "mcpServers": {
    "findleademails": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://www.findleademails.com/api/mcp",
        "--header",
        "Authorization:${AUTH_HEADER}"
      ],
      "env": {
        "AUTH_HEADER": "Bearer YOUR_API_KEY"
      }
    }
  }
}

From a shell, with the Claude Code CLI installed:

claude mcp add --transport http findleademails https://www.findleademails.com/api/mcp --header "Authorization: Bearer YOUR_API_KEY"

--transport http is the remote HTTP transport, and --header sets the bearer key. Replace YOUR_API_KEY before you run it.

Workspace config is .vscode/mcp.json. VS Code's file uses a top-level servers object and requires type "http" for a remote server.

{
  "servers": {
    "findleademails": {
      "type": "http",
      "url": "https://www.findleademails.com/api/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

VS Code's docs prefer an input variable over a key written into the file. The block above uses YOUR_API_KEY so the header is obvious; swap in their input form if you will commit the file.

A client that takes a remote server as a URL plus headers uses this shape. If that client's own docs use different keys, follow those docs. We have checked this file against Cursor; we have not checked a separate config format for every other product.

{
  "mcpServers": {
    "findleademails": {
      "url": "https://www.findleademails.com/api/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

This calls the tools/list method. The handler requires both Accept values on a POST. A success comes back as a server-sent event: message whose data line is the JSON-RPC result listing the five tools. No session header is required.

curl https://www.findleademails.com/api/mcp \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'

A missing or rejected key stops first, with HTTP 401 and {"error":"invalid_token","error_description":"No authorization provided"}. Nothing is listed and nothing is charged.

Each tool takes the same fields as the matching REST route. A single lookup wants first_name, last_name, and company_domain. A LinkedIn company page URL can stand in for the domain; resolving it costs 1 credit only when it finds a match, separate from the lookup.

find_emails_bulk is synchronous and capped. Above that cap it returns batch_too_large with max_rows. get_job_results pages with cursor until next_cursor is null, and it can be read while the job is still running.

Billing is the same wallet as the dashboard and the REST API. A verified result costs 1 credit. A catch-all is free — the domain accepts every address, so no specific mailbox can be confirmed. A miss (not_found) is free, and so is a repeat lookup of someone already found. A predicted result is free when verification ran but could not reach a verdict.

create_job reserves one credit per row up front and refunds the unused portion as rows resolve. Credits never expire.

find_email, find_emails_bulk, and create_job use the same per-workspace ceilings as the synchronous REST endpoints. The numbers are on Rate limits. Reading a job's status or results is not rate limited.

A failed tool call is not an HTTP error from the agent's point of view. The tool result is flagged as an error and its text is JSON { "error", "code", ... }.

Before a tool runs, a missing or rejected key is HTTP 401. After a valid key, a POST without Accept: application/json, text/event-stream is HTTP 406.