how-to · Settings Open this in the console

Connect an AI agent to BE Hub (MCP).

Settings › connect an AI agent. The address is your hub plus /mcp. The agent authenticates with a user's API key.

The agent can do exactly what the user behind the key can, in that user's companies: read machines, alerts, events, patches and BE AI runs; ask BE AI; acknowledge, comment on and resolve alerts; change a machine's alert settings and reboot windows; approve or reject BE AI proposals; run saved scripts; work tickets (raise, reply, assign, log time, accept BE AI's triage, mark spam) and the knowledge base; read security findings, accept, assign or mark them done, check a machine now or leave it out; propose and approve group security fixes; run any report; and see what is pending across its companies. It cannot open the screen, files or a terminal; those stay with a signed-in person. Every tool call is in the audit trail as mcp_call, under that user.
01

Copy the address

Open Settings. Connect an AI agent shows this hub's MCP address and the line that adds it to Claude Code. Replace bek_... with the API key of the user the agent should act as. Any MCP client that speaks streamable HTTP works the same way: the address, and an Authorization: Bearer header with the key. Start the agent with whoami and list_machines.

Connect an AI agent box with the MCP address and a claude mcp add command Connect an AI agent box with the MCP address and a claude mcp add command
The MCP address and the Claude Code line, each with a copy button.

Make a user just for the agent so its actions are easy to tell apart: Invite a teammate shows the key once. A key can be replaced at any time: Replace a script key. An operator user cannot manage users or billing, which is usually what an agent should have.