As of Oct 2026, CLI personal access tokens support editor/manager roles for write access but don't require agents to provide reasons for each credential access. Meanwhile, Web UI AI tokens force reason logging but are read-only only.

Proposal: Add a --agent flag to pat access grant: pass-cli pat access grant --pat-name "AI Agents" --vault-name "Mimir" --role editor --agent When --agent is set, every credential retrieval via that token requires the caller to pass a --reason parameter, matching Web UI behavior while preserving write capability. This enables secure CI/CD pipelines where audit trails matter, not just home automation scripts.