Connect your agent

• A Cornilius account. Sign up takes an email address and a Microsoft sign-in.
• An MCP-capable agent. Claude.ai is verified today; any client that can call an MCP server over HTTP follows the same pattern.

https://api.cornilius.ai/api/mcp

Every connection uses this URL. What differs per agent is how it authenticates: a guided approval flow, or an API key.

Claude.ai authenticates through the web: give it the server URL and it opens a Cornilius approval screen in the browser. No key is pasted anywhere.

1. 01

   Add the connector

   In Claude.ai, open Settings, then Connectors, then Add custom connector. Paste the server URL above.

2. 02

   Approve access

   Claude.ai opens a Cornilius approval screen. Sign in with your Cornilius account and approve. The connector is now linked to your workspace.

Claude Code, and most other MCP clients, authenticate with an API key sent as a bearer header.

1. 01

   Create a key

   Sign in and open Connection, then select Create API key. Copy the key — it is shown exactly once. If it is lost, revoke it and create a new one.

2. 02

   Configure the client

   In your client's MCP configuration, set the server URL and send the key as a bearer header:

   Authorization: Bearer crn_YOUR_KEY

3. 03

   Confirm the key works

   To test the key before involving the agent:

   curl -s https://api.cornilius.ai/api/mcp \
     -H "Authorization: Bearer crn_YOUR_KEY" \
     -H "Content-Type: application/json" \
     -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

   A working key returns the tool list. An invalid or revoked key returns 401.

Cornilius ships a skill file that teaches your agent the command vocabulary — listing playbooks, executing them by id, recalling past findings, and escalating a recurring issue to its recorded owner. Download it from the Connection page, or fetch it directly:

https://cornilius.ai/skills/claude/cornilius.skill.md

In Claude.ai: Customize, then Skills, then Upload a skill. The connection works without it; the skill makes the agent use it well.

Ask your agent to start a Cornilius session. It calls the cornilius_start tool and reports your workspace status. On the dashboard, Usage shows the connection as active after the first successful call.

Teach it your business: encode your metric definitions and playbooks so the agent works your way, in every conversation.