agent-chaperone
Screens an AI agent's tool calls before they run and tool results before the agent reads them. An MCP proxy plus a hooks adapter for a client's built-in tools.
How it uses Jev
Set TYPESAFE API KEY for the model's judgments; without it only the deterministic rules run, and the log says so. Then use the agent as you normally would, and read what it would have done:
Highlights
- Holds a dangerous call before it runs. Destructive commands, data leaving the machine, a credential in the arguments. On Claude Code a held call becomes the permission prompt you…
- Withholds a result written to instruct the agent. A result that scores high enough is replaced with a notice. One that scores lower reaches the agent with a warning marking the su…
- Keeps secrets in. Secret-shaped strings are redacted before anything is sent to be screened or written to the log, and a call that looks like it sends private data out is held.
Getting started
On Claude Code, the plugin covers the client's own shell, edits and fetches. From inside a session:
Source: community catalog. Details and metrics reflect the published community snapshot and are not a performance endorsement. Documentation details extracted from GitHub · the README · the project website.