Bring robot regression testing into your development workflow. Authenticated pipelines can request evaluations, read evidence and stop for human approval.
Reuse robot Motion Sessions, ingest telemetry, then call MotionBench with a frozen reference and the candidate from the pipeline.
GitHub Actions can request MotionBench with a service-account API key and record the evaluation id. MCP clients can call comparison tools. ROS2 bags convert through the offline ros-bridge before evaluation.
Pipelines produce evidence for people. Release decisions remain human-gated.
| Surface | Availability | Notes |
|---|---|---|
| Motion API evaluations | Available | Workspace-scoped API keys |
| MCP tools | Available | Same scopes and audit as REST |
| GitHub CI evidence ids | Available | request_id on audit events |
| Agent interoperability | Research preview | Expanding with authenticated testing |
MotionBench evaluations can run from authenticated Motion API or MCP tools in CI pipelines. Human review stays required before release. Start at /robot-ci-cd.
Use authenticated Motion API or MCP tools to compare reference and candidate sessions, then keep human review before release.
Evaluation pipelines support evidence export and human review. Release decisions stay with your team.
Loading Arisole. If the interactive app does not start, continue with the public links above.