Integrations for the stack you already run

Run your eval suite on every pull request and block the merge when scores drop.

Merge request checks for teams on GitLab.com or a self-managed instance.

Test preview deployments against real cases before they get promoted.

Post run results, blocked merges and drifting production scores to a channel.

Turn failing cases into issues with the input, output and grader note attached.

Send grader scores and cost per request as metrics next to your latency graphs.

Pull cases straight from a read replica with a SQL query you control.

Capture traces from LangChain apps with a callback handler, no code rewrite.

Link model errors and timeouts to the prompt version that caused them.

Run any OpenAI chat or reasoning model in your suites with your own key.

Test Claude models side by side with your current provider.

Try open-weight and hosted Mistral models against your golden set.