Skip to content

Jentic + Claude

Goal — Give Claude two Jentic MCP servers: the hosted directory for natural-language discovery of API operations before you've installed anything, and your self-hosted Jentic One as a custom connector for the full search → inspect → execute loop against your registry.

New to Jentic One? Start with the MCP overview and the Quickstart.

1. Hosted directory search (discovery)

Point Claude at the hosted MCP server to search the public Jentic API directory in natural language — "find an operation that creates a GitHub issue" — and get back concrete operations before any deployment exists.

Easiest — the Jentic connector (Anthropic-verified). Jentic is published in Claude's connector directory, so there's nothing to paste:

  1. Open the Jentic connector and click Add to Claude (or in Claude, go to Settings → Connectors → Browse Connectors and search for Jentic).
  2. Sign in when prompted.

That connects the hosted https://api.jentic.com/mcp server, exposing search_apis, load_execution_info, and execute against the public directory.

Any other client — add it by URL. Point a URL-capable client at the same endpoint:

{
  "mcpServers": {
    "jentic": {
      "url": "https://api.jentic.com/mcp"
    }
  }
}

No token to configure — the hosted server uses OAuth (Dynamic Client Registration), so the client registers itself and prompts you to authorize on first connect.

This surface is for discovery before you've installed anything — which API and which operation solves the task. For the full loop against your own registry, connect a self-hosted instance below.

2. Self-hosted Jentic One (full search → inspect → execute)

A self-hosted instance gives Claude the full loop against your own registry — search your imported operations, inspect a contract, then execute it — with credentials decrypted only inside your Broker. Give Claude access one of two ways.

If the jentic CLI is on the machine Claude runs on, jentic setup registers the agent and writes the Jentic skill into Claude Code's native layout:

jentic setup --operator claude

It registers the agent identity, prints an approval link and waits for an operator to approve it at /app, mints tokens, and installs a spec-conformant SKILL.md. Re-running is idempotent. Already registered? Install just the skill:

jentic skill init --operator claude

Once installed, Claude drives Jentic on its own — ask it in natural language and it calls search_apis → inspect_operation → execute for the operations your agent has been granted, via the local jentic mcp stdio server.

URL connector (hosted /mcp endpoint)

If your operator has enabled the deployment's hosted /mcp endpoint (server.mcp.enabled: true), Claude can connect over a URL with no CLI on the machine. How it authenticates depends on the deployment:

OAuth (no token to paste) — if the operator also set server.mcp.oauth.enabled: true, the endpoint uses OAuth with Dynamic Client Registration, like the hosted directory. Claude registers itself and prompts you to authorize on first connect (the self-registered client waits for operator approval unless auto_approve_clients is on):

{
  "mcpServers": {
    "jentic": {
      "url": "https://<your-jentic-one-host>/mcp"
    }
  }
}

Manual bearer — with only server.mcp.enabled: true (OAuth off), supply the agent's bearer per request:

{
  "mcpServers": {
    "jentic": {
      "url": "https://<your-jentic-one-host>/mcp",
      "headers": { "Authorization": "Bearer <agent-api-key>" }
    }
  }
}

The bearer comes from jentic register. The agent detail page's MCP tab renders the right snippet pre-filled for whichever mode the instance serves.

Support