Skip to main content

Connect InNotes to your AI

Add the InNotes MCP server to Claude, Cursor and other AI assistants so they can remember and recall facts about your contacts.

InNotes exposes an MCP (Model Context Protocol) server. Once connected, your AI can store memories about the people in your network and recall them later to help you keep relationships warm.

What you can ask once it is connected

These all work out of the box on a connected account:

  • “Remember that Anna Rossi just moved to Zurich and is looking for a senior backend role.”

    Stores a memory on the contact (remember).

  • “What do I know about Marco Bianchi before my call tomorrow?”

    Pulls everything on one person: memories, notes and profile (recall, search_notes, get_contact).

  • “Who in my network works in fintech and sails?”

    Semantic search across every memory you have stored (recall).

  • “Add a note on the Acme Corp job: they asked for a take-home exercise.”

    Writes a note on a saved job (list_jobs, create_note).

  • “Which of my follow-ups are overdue?”

    Lists time-bound memories whose date has passed (list_memories).

MCP endpoint

https://innotes.me/api/mcp

Get an API token (for token-based clients)

Some clients connect with a token instead of a login. Create one under Profile → API Tokens and use it in the Authorization header shown below. You can revoke it anytime.

Open API Tokens

Add it to your AI

Claude (claude.ai)

Open Settings → Connectors → Add custom connector, paste the endpoint above and confirm. Claude sends you through the InNotes login and consent screen — no token needed.

Claude Desktop

Settings → Connectors → Add custom connector, then paste the endpoint. Sign in to InNotes when prompted.

Claude Code / CLI

Create an API token, then run:

claude mcp add --transport http innotes https://innotes.me/api/mcp \
  --header "Authorization: Bearer inn_..."

Cursor

Add an MCP server in Cursor's settings (or ~/.cursor/mcp.json) with the endpoint and your token:

{
  "mcpServers": {
    "innotes": {
      "url": "https://innotes.me/api/mcp",
      "headers": { "Authorization": "Bearer inn_..." }
    }
  }
}

ChatGPT & other MCP clients

In any client that supports custom MCP servers, add a Streamable HTTP server pointing at the endpoint, with your token in the Authorization header. Where MCP connectors live — and whether they are available — varies by client and plan.

What the connector can do

Tools, by group

Every tool acts only on your own InNotes account. Read tools are marked read-only, so your assistant can run them without asking each time. Tools that change or delete something you already have are marked destructive and always ask first. Creating a new note, contact or memory is marked additive, so depending on your client it may go through without a prompt.

  • Memories — remember, recall, list_memories, update_memory, delete_memory. Durable facts about the people in your network, always private.
  • Notes — list_notes, get_note, create_note, update_note, delete_note, search_notes. Your notes on contacts and jobs.
  • Contacts — list_contacts, get_contact, create_contact, update_contact, delete_contact.
  • Jobs — list_jobs, get_job, create_job, update_job, delete_job, plus list_statuses and list_tags for the valid status and tag values.

Getting the most out of memories

  • Anything that is only true until a certain date — an open action item, a deal stage, an availability window — is worth storing as a “time_bound” memory with a date. Past that date it still comes back, flagged as no longer confirmed, instead of quietly going stale.
  • Asking your assistant for overdue follow-ups every so often keeps the store trustworthy: confirm what actually happened, then either push the date out or delete the memory.
  • Recall matches on meaning rather than words, and works best when the store is written consistently in one language. If you work across languages, telling your assistant to keep memories in English gives the most reliable results.
  • Your saved InNotes data answers most questions on its own. The live LinkedIn tools are for what is not saved yet: finding new people, reading a full career history, or checking messages and notifications.

Your data stays yours

The connection only ever acts on your own InNotes account, and AI memories are always private. Revoke access anytime from Profile → API Tokens, or by removing the connector in your AI client.

What we record about connector usage — counts of which tools are called, never their contents — is described in section 12.2 of our privacy policy. Read the privacy policy · Terms & conditions · Support: info@innotes.me