Ship integrations that agents get right.

Your users’ agents write the integration code now. Run them through real tasks with your API, SDK, CLI, or MCP server, grade the result, and catch regressions before release.

Real tasks, your criteria

Give agents the integration work your users ask for, then grade the finished code against the criteria you define.

Every surface you ship

Test SDKs, CLIs, native binaries, and authenticated MCP servers and APIs in isolated sessions.

Checks on every release

Run evaluations on pull requests with the Gauge GitHub App, or trigger hosted runs from your own CI.

FAQs

Yes. Choose the repository commit, harness, model, skills, and tools for each run, then repeat the same task to measure the effect of a change.

Yes. Start from a repository at a pinned commit, including private GitHub repositories in hosted Gauge.

Yes. The Gauge CLI includes instructions for coding agents, and the Agents API lets you launch evals and read results from your own tools.