API & MCP reference

LinkedMinds exposes a Model Context Protocol (MCP) server over HTTPS. Any MCP client can use it - Claude Desktop, claude.ai, Claude Code or your own - by sending a Bearer token. Generate credentials from Settings → API Key inside your dashboard.

Add the server to Claude in three steps.

1. 1Go to Settings → API Key, open Claude MCP Credentials, and click Generate. Copy the Client Secret - it's shown once.
2. 2In claude.ai → Settings → Connectors (or Claude Desktop → MCP Servers), choose Add custom MCP server.
3. 3Enter the URL below and your credentials, then save.

Claude Desktop config.json alternative:

{
  "mcpServers": {
    "linkedminds": {
      "type": "streamable-http",
      "url": "https://mcp.linkedminds.in/mcp",
      "headers": { "Authorization": "Bearer <YOUR_CLIENT_SECRET>" }
    }
  }
}

Once connected, ask Claude in plain language - it calls these under the hood.

For your own code rather than a chat client, there is a JSON REST API on the same host. It uses a separate credential - the REST API Key in Settings - and is available on the Pro and Enterprise plans.

curl https://mcp.linkedminds.in/v1/campaigns \
  -H "Authorization: Bearer lm_YOUR_REST_API_KEY"

Replies and withdrawals are metered by the hour rather than the minute, because they act on a real LinkedIn account. Those are ceilings on the API, not permission to send: every queued action still passes the same daily caps, warm-up phase and ban-risk checks as the dashboard - the API is a second door to the product, not a way around it. Full reference, including every field, header and error code, is in the dashboard under Docs.

Your own leads go in over the same API. POST /v1/campaigns/{id}/leads takes up to 200 LinkedIn profile URLs a call, which is what turns a sheet, a CRM or a scraper into a running campaign. Re-sending the same list is safe: anyone already on the campaign is skipped, never restarted. A campaign holds up to 10,000 leads on Pro and 50,000 on Enterprise.

n8n, Zapier and Make need nothing special. There is no custom node to install: point an HTTP Request node at the endpoint and send the key as a Authorization: Bearer header, stored as a credential rather than inline in the workflow. Poll on a five-minute schedule or slower, and honour Retry-After on a 429. The step-by-step version is in the dashboard docs.

If a key looks compromised, we act on it. Both credentials are watched. A run of wrong secrets against your key means someone has seen part of a real one, so we disable it and tell you - by email, in the app, and on the Settings page, with the reason. Generating a new key clears it and nothing else about your account is touched. One valid key being used from an unusual number of places gets you a warning but is never disabled automatically: that is also what a normal serverless integration looks like, and breaking a working automation on a guess would be the bigger harm.

Driving LinkedMinds from Claude changes nothing about how carefully it acts. Every action runs through the same pacing engine as the dashboard: randomised delays of 2-10 minutes, burst windows with cool-down breaks, a four-week warm-up on every new account (5 requests a day in week 1, rising to 20 by week 4), and per-account caps set at or below LinkedIn's own.

Database searches never contact LinkedIn, so they carry no ban risk and do not spend your LinkedIn search quota. Full parameter-level reference for all 22 tools lives in the dashboard, under Docs.