Discovery is where you see how agents experience a service you’ve added, and fix it. It leads with the agent experience score: the same check, five stages and score as the public agent experience report, so the number in Forge, on your report link and in these docs is always the same. Beside it, Copy the fix prompt copies one prompt that applies every fix. Paste it into Claude Code, Cursor or Codex in your repository, then select Re-scan.

​
Things to fix

Every problem Forge found, biggest first, with its stage (Reach, Navigate, Understand, Use or Pay) and the points it adds to your score. Select one to see what it means for an agent and the exact change to make. How each stage is checked and scored is on the agent experience page.

​
What already works

Everything that passed, in one line you can open into the full list, including standards Forge found such as OpenAPI, MCP, SKILL.md and x402. Each still opens to say what it means. Stages that don’t apply to your product are listed underneath and left out of the score.

​
How agents get around

At the bottom, the map shows how agents get around your service: your website, the hosts it leads to (your API, docs and MCP server), and what agents reach on each, such as pages, API routes, paid routes and MCP tools.
  • Hosts and lines in red have something to fix. Click one to open its fix above.
  • Point at a host to light up its points, or at a point to light up its host.
The map has no score of its own: the agent experience score is the one number to follow.

​
Checks and re-scans

Discovery shows your service’s latest check. If there isn’t one yet, it runs one when you open the page, with the check animation while it works (it can take up to a minute). Re-scan, at the top right of the page, runs a new one, for example after you ship a fix.