Configuration

Agent endpoint

Any HTTP URL that accepts POST and returns a response. Works with custom agents, OpenAI-compatible APIs, and any framework behind an HTTP wrapper.

AI Judge (per-project)

Set a judge endpoint in the project's Judge tab — any OpenAI-compatible URL + model name + API key. The key is encrypted at rest (AES-256-GCM). Leave blank to use the server default.

Categories

Tag test cases with categories like access_control, audit_logging, human_oversight to map them to compliance controls (HIPAA, EU AI Act).

Context (RAG)

Add a context field with source documents. The judge runs an additional faithfulness check — is every claim grounded in the context?