Directory / Inspect
Inspect
Verified
UK AISI open framework for evaluating language-model agents.
Inspect is an open-source evaluation framework from the UK AI Security Institute. It supports model and agent evaluations, including sandboxed tasks. A separate tool handles pull-request commenting.
- Model support
- Multi-model
- Deployment
- Self-hosted
- Open source
- Yes
- License
- MIT
- Pricing model
- Free
- Added / updated
- 19 Aug 2026 / 19 Aug 2026
inspect.aisi.org.uk · Docs · Source
Assessment
Inspect suits rigorous agent evaluations where task design and reproducibility matter. Its footprint exceeds Promptfoo; pull-request comments and merges require another tool.
Strengths
- Public evaluation framework with sandbox-aware tasks.
- MIT, model-agnostic.
- Agent-focused evaluation alongside chat-prompt tests.
Limitations
- Steeper than product eval tools.
- No GitHub app.
- Task design remains the operator's responsibility.