Library
eg
Agents & MCP · 2026
evalgate
a safety check for AI assistants that catches bad behavior before it ships
record once replay as a repeatable test
every change checked before it goes live
v0.1 first release, building in the open
recorded session, step by step
replay run vs the saved baseline
checks that passed or failed
About
A tool that turns a real AI assistant session into a repeatable test. It records what the assistant actually did, step by step, then replays that run later to confirm the behavior still holds. When someone tweaks the assistant, this catches changes that quietly break it before they reach real users.
The idea: teams can watch what their AI does, but most have no way to fail a build when an edit makes it misbehave. evalgate is the missing safety check. The headline example wires it into a trading assistant, where rules like "never place an oversized order" become hard checks that stop a risky change before it can ever reach the broker.
Liner notes
- Records a real session, then replays it the same way every time.
- Checks what the assistant did, not just what it said.
- Built in the open on an eleven-day sprint to a first release.
Credits
Language python
Shape safety check for AI assistants
Example trading assistant guardrails
Status building → v0.1