Skip to content

Connect Cursor to Sensefold

Part of Sensefold for agents — the overview, permission tiers, and tool table.

Cursor supports remote MCP servers with OAuth. The normal setup is a URL-only entry plus a browser sign-in; no key is stored in any file.

  1. In Cursor, open Customize → MCPs → New MCP server. Cursor opens mcp.json in the editor.
  2. Add this entry (inside the existing mcpServers object) and save:
    "sensefold": {
    "url": "https://api.sensefold.app/mcp"
    }
  3. Back in Customize → MCPs, Sensefold shows a login prompt — click Connect, sign in in the browser, and pick a permission tier: read-only (search and read), edit (also save and update), or full (also delete). Start with read-only unless Cursor needs to save or update items.
  4. Sensefold now appears under Connected with a green dot and its tool count — 6 tools for read-only (including the search/fetch compatibility aliases), 10 for edit, 11 for full.

The connection appears under Sensefold → You → For agents, where you can revoke it anytime.

If a browser sign-in is not possible, use an Agent key from Sensefold You → For agents and reference it through an environment variable — never as plain text in the file:

"sensefold": {
"url": "https://api.sensefold.app/mcp",
"headers": {
"Authorization": "Bearer ${env:SENSEFOLD_AGENT_KEY}"
}
}

Keep keys out of git: prefer the user-level ~/.cursor/mcp.json over a project-level .cursor/mcp.json. If a key leaks, revoke it in Sensefold under You → For agents and create a new one.

  • “Search my Sensefold library for the architecture notes about the sync layer.”
  • “Save a note: decisions from today’s refactor.”

See the MCP tools reference for everything Cursor can call, agent permissions for tiers and undo, and troubleshooting if the connection fails.

Tiers, revocation, and undo: Permissions. What the client receives: Privacy for connected AI. Problems: Troubleshooting.