Connect Claude Code to Sensefold

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

Claude Code supports remote MCP servers with OAuth, so the normal setup needs no key: one command, then sign in in the browser and pick a permission tier.

Setup

Add the server:

claude mcp add --transport http sensefold https://api.sensefold.app/mcp

Inside Claude Code, run /mcp, select sensefold, and choose Authenticate. Sign in in the browser and pick a tier — read-only (search and read), edit (also save and update), or full (also delete). Start with read-only unless Claude Code needs to save or update items.
Run /mcp again: Sensefold shows as connected with its tool count — 6 tools for read-only (including the search/fetch aliases), 10 for edit, 11 for full.

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

Headless / CI

Where no browser can complete the OAuth flow, use an Agent key instead:

Create a key at Sensefold You → For agents, pick a tier, and copy it — the full key is shown only once; if it is lost, revoke it and create another. Never paste it into a chat.

Add this to the project’s .mcp.json (or the user-scoped equivalent):

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

Set SENSEFOLD_AGENT_KEY in the environment that launches Claude Code, and run claude mcp get sensefold to verify.

Try it

“Search my library for prior decisions about credit pricing before we design this.”
“Save a note summarizing what we shipped today.”

Every write is versioned and revertible — see agent permissions. An agent-readable version of this setup lives at /for-agents/skill.md — you can paste that URL to Claude Code and let it configure itself. Connection problems: troubleshooting.

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