Skip to main content

​
Finding your way around EvalGate

The EvalGate web app has two main areas: the sidebar on the left and the main panel on the right. Everything you do happens in one of them. A screenshot of the EvalGate workspace showing the sidebar on the left and the main panel on the right.

​
The sidebar

The sidebar is how you move between features. It is shaped as a loop — observe, measure, improve, ship — with a few pinned destinations above it.
  1. Pinned — Release readiness (/agent-evidence), Review, and Costs.
  2. Observe — Traces, Logs Explorer, Trajectories & Workflows, Production Insights, and Dashboards.
  3. Measure — Evaluations, Eval Packs, Knowledge Network, Datasets, Scorers & Judges (Scorers, LLM Judge, Calibration, Annotations), Measurement, and Benchmarks.
  4. Improve — Experiments, Repository AI Map, Playground (Playground and Prompts), and Red Team.
  5. Ship — Verification (/eval-loop), Golden Release, and Continuous Eval.
  6. Developer — API keys, remote runners, deployable assets, and asset portability.
Copilot lives on the dashboard and on Prompt, Dataset, and Scorer hubs, plus /copilot for proposal history. The workspace switcher at the top of the sidebar changes workspaces, opens settings, and signs out. There is no mounted command palette (Cmd+K) and no Records chrome section. If the sidebar is in your way, click the collapse icon at the top to hide it.

​
The main panel

The main panel shows the content for whatever you selected in the sidebar. Use the page’s own actions — New evaluation, New prompt, Bind to evaluation — rather than a global overflow menu.

​
Where each feature lives

​
Up next

Make your first local gate fail in CI.