Skip to main content
PeerNotes exposes your workspace as a remote MCP server. Point the AI agent you already use at it and, instead of guessing from the open web, it answers from what you and your teammates actually wrote — your notes, sources, and topics.
“What did the team learn about screening signal this quarter?” — Read 9 notes and 4 sources from your workspace. The recurring theme is a synthesis gap, not a capture problem…
Using Claude Code or Claude Cowork? Install the Claude plugin instead. It connects to the same MCP server, and adds skills and scheduled syncs for meetings, code, and documents. Use the steps on this page for other agents.

​
Connect it

1

Copy your workspace address

https://api.peernotes.io/mcp
2

Add it as a connector

Paste the address wherever your agent asks for a remote / HTTP MCP server or connector.
3

Sign in with PeerNotes

Authenticate with your PeerNotes account when prompted — no API keys, no files to edit.

​
Or edit a config file

For clients that take a JSON config, add:
The same server works in Claude without the plugin. In Claude Desktop, open Settings › Developer › Edit config, then restart Claude. In Claude Code, run:
Full instructions live at get.peernotes.io/download.

​
Works with

  • OpenAI Codex, GitHub Copilot, and Cursor
  • Claude on web, desktop, and mobile, as a custom connector
  • Any other client that supports remote MCP servers
MCP is an open standard, so the list keeps growing on its own.

​
What your agent can see

Your agent only ever reaches the workspaces your PeerNotes account can already open — nothing else, and nothing that isn’t shared with you. Access follows your account: leave a workspace and the connection to it goes with you.
  • Claude plugin: the recommended way to use PeerNotes from Claude Code and Cowork
  • Chat: the Peer Agent, inside PeerNotes itself
  • Peer Agent: how PeerNotes’ own AI is grounded