{"id":"inspect-ai","name":"Inspect AI","maker":"UK AI Security Institute","url":"https://inspect.aisi.org.uk","repo":"UKGovernmentBEIS/inspect_ai","package":"pypi:inspect-ai","summary":"A Python framework for writing and running evaluations of language models and agents, from the UK AI Security Institute.","details":"Defines evaluations as tasks with datasets, solvers and scorers, supports agent scaffolds, tool use and sandboxed execution, and has a log viewer. A companion collection of ready-made evaluations is maintained separately. The framework is open source under the MIT licence.","open_source":true,"hosted":false,"speaks":["mcp"],"also":["benchmarks"],"verified":"2026-11-15","category":"observability","category_name":"Observability and evaluation","page":"https://modelmoot.com/tools/inspect-ai","reviewed":"2026-11-15","checked":{"at":"2026-11-15T14:03:44.997Z","address_answers":true,"repository":{"at":"2026-11-16T00:27:04.284Z","archived":false,"licence":"MIT","language":"Python","last_change":"2026-11-15T22:52:40.000Z","latest_release":{"tag":"0.3.290","published":"2026-11-14T21:14:27.000Z"}}},"note":"Descriptions are the site's own. Facts under \"checked\" were read automatically from the tool's address and public repository."}