Skip to main content
AO exposes your infrastructure to AI assistants through the Model Context Protocol. Every tool call is scoped to your organization and recorded in the audit log. Claude launches the local AO agent over stdio. The agent connects upstream to https://mcp.automatedoperations.com, owns the refreshable session, and keeps separate account and organization profiles. The first use opens a browser to sign in and pick an agent role; switching to an existing profile does not require another login. Agent roles are a separate, org-wide pool curated under IAM → Roles → Agent roles — distinct from the user roles assigned to members. Any member can pick any agent role for a session, and the agent’s effective access is always capped at your own. The catalog owner controls the menu; you choose how much of it to lend the agent.

​
One-command setup

This detects the AI clients installed on your machine (Claude Code, Claude Desktop, Cursor, VS Code, and more — all pre-selected, deselect any you don’t want), configures each one to launch @ao-mcp/local over stdio, writes a backup, and verifies the connection. Needs Node 20+. The rest of this page is the manual equivalent for Claude specifically. For other clients, see the client matrix.

​
Claude Code

Then run /mcp inside Claude Code to complete the sign-in on first use.

​
Claude Desktop

The setup command writes the same stdio entry to Claude Desktop’s MCP config:
Restart Claude Desktop after changing its configuration. The local AO agent prompts you to authorize with your AO account and select an agent role.

​
What your assistant can do

  • List services and hosts, with live status
  • Pull recent changes and incident-relevant context
  • Run on-demand diagnostics against a host — fresh answers, not stale dashboards
Every call is permission-checked and lands in your audit trail.