Forge runs an MCP server, so the agent you already use (Claude, Cursor, Codex, and others) can read the same data you see in the dashboard, and run a check when you ask.
Server URL
The first time your agent connects, a browser window asks you to sign in to Forge and approve access. Your agent sees only the services in your account.

​
Connect your agent

Connect agent in the Forge sidebar has the same instructions for more clients, including Hermes, GitHub Copilot and OpenCode.

​
Try asking

  • “What needs my attention across my services this week?”
  • “Which of my routes dropped in the rankings, and who passed them?”
  • “Why is the weather route failing its health check, and how do I fix it?”
  • “Which agents paid for my API in the last 30 days, and how did they rate it?”

​
Tools

The last four are available when you grant Forge’s forge:operate permission at sign-in. They change nothing about your service; they queue the same scans as the buttons in the dashboard.
This is different from the docs MCP, which lets an agent read this documentation.