Cassette records your app's LLM & agent API calls, then replays them in CI — so tests stop being slow, expensive, and flaky. The team tier catches agent behavior regressions on every PR.
Free & open source. Records to local files; no gateway, no infra.
pip install cassette-sdk # Python
npm install cassette-sdk # Node / TypeScript
from cassette.recorder import http_client
from openai import OpenAI
client = OpenAI(http_client=http_client(project="demo")) # records → replays
First test run saves each LLM response to a portable cassette file.
Every run after replays it — instant, $0 tokens, no flakiness.
On each PR, the Team gate flags when your agent's behavior actually changed.
The recorder + replay, local cassettes, unlimited. For individuals & open source.
InstallShared registry + the GitHub PR behavior-gate on semantic drift. Storage included.