Jev in production › Agent guardrails and approvals
OpenCode and Claude Code plugin that sends unresolved shell permission requests to Jev as four Noul questions (harmful, sensitive, untrusted, obscured) and auto-approves only when all pass calibrated thresholds; otherwise the normal prompt runs.

The permissions plugin auto-approves harmless shell permission requests. Anything uncertain continues through your agent's normal permission flow.
In this isolated demo, a test reviewer requests the nonmatching OpenCode approvals.
Claude checks readable ask settings; managed, command-line, and session-only rules are not visible to its hook.
Starting in 0.6.0, the OpenCode integration in Jevvy Permissions requires OpenCode 2.0.27 or newer.
Starting in 0.9.0, Jevvy Permissions also provides a Jevvy indicator in the session header on OpenCode Desktop 2.2.0 or newer. Its text is green when auto-review is enabled, red when disabled, and neutral when status or setup is unavailable. Click it or use Toggle Jevvy Permissions in the command palette to...
For the project's own README, linking back here: