Connect your AI (MCP setup)

Point Claude, Codex, Gemini, or any MCP client at your Banter workspace — your AI, your subscription, your notes.

MCP (Model Context Protocol) is the open standard that lets AI tools work with outside apps. Connecting Banter over MCP means the AI you already use — Claude, Codex, Gemini, Cursor — can search your notes, read transcripts, check your todos, and file new notes, all from its own chat window, on your own subscription.

What you need

  • The server address: https://api.usebanter.ai/mcp
  • A token (for most clients): in Banter, open Settings → Connections and click Copy token. Treat it like a password — it grants access to your workspace.
  • A test: once connected, ask your AI “List my Banter folders.” If it answers with your folders, you’re done.

Claude is the exception — it signs in instead of using a token.

Claude (easiest)

  1. In Claude, go to Settings → Connectors and click Add custom connector.
  2. Paste https://api.usebanter.ai/mcp as the server URL.
  3. Click Connect and sign in with the same Google account you use for Banter. No token needed.
  4. In a new chat, enable the Banter connector and ask: “List my Banter folders.”

Codex

Codex reads the same configuration from its CLI and IDE extension, so adding it once covers both:

export BANTER_MCP_TOKEN="<paste token from Banter Settings → Connections>"
codex mcp add banter --url https://api.usebanter.ai/mcp --bearer-token-env-var BANTER_MCP_TOKEN
codex mcp list

Restart Codex if it was open, then ask it to list your Banter folders.

Gemini CLI

export BANTER_MCP_TOKEN="<paste token from Banter Settings → Connections>"
gemini mcp add --transport http banter https://api.usebanter.ai/mcp \
  --header "Authorization: Bearer ${BANTER_MCP_TOKEN}" \
  --trust
gemini mcp list

Add --scope user if you want the connection available in every project.

Any other MCP client

For clients configured with a JSON file (Cursor and many others), use this shape:

{
  "mcpServers": {
    "banter": {
      "url": "https://api.usebanter.ai/mcp",
      "headers": {
        "Authorization": "Bearer <YOUR_BANTER_MCP_TOKEN>"
      }
    }
  }
}

If it doesn’t connect

  • 401 / unauthorized — the header must be exactly Authorization: Bearer <token>, and the token must be freshly copied from Settings → Connections.
  • Claude doesn’t show Banter — after adding the connector, you still need to enable it in the chat’s tools menu.
  • The AI can’t find something — tell it to list folders first, then search. Like a new colleague, it navigates better once it’s seen the layout.

What now?

See what connected agents can do for the full picture of what your AI can read and write, and MCP recipes for the workflows this unlocks — bulk imports, workspace cleanup, and having your AI write its work into Banter.