Gauge released an open-source CLI for improving Agent Experience. It runs real agents against a product to find where they get stuck. The tool captures friction points like blocked commands or unused Skills. It lets a coding agent test fixes before publishing them. The loop runs locally with Docker and existing agent subscriptions. No Gauge account is required for local use. It supports Claude Code, Codex, Pi, and OpenCode. Users can define tasks, test current products, and analyze results. Fixes can be validated before release. The CLI also supports hosted execution with managed infrastructure.