Gensee CrateFull-stack, long-horizon runtime safety
for AI coding agents.

Gensee Crate watches system events,
requests, tool calls, skills, and memory
behind Claude Code and Codex.

Observe What Agents Do

Capture prompts, tool intent, commands, files, network targets, alerts, and timeline context in a local store.

Run the sidecar

Apply configurable allow, ask, and deny decisions for secrets, destructive operations, control-plane writes, and suspicious artifacts.

Read the policy model

Use managed macOS sandboxing and staged workspace review when you want Gensee to launch the agent.

Use gensee run

Wire Claude Code or Codex into the same timeline so tool requests and runtime effects are evaluated together.

Configure hooks

Link prompts, tool calls, file effects, artifacts, alerts, and review verdicts for post-run inspection.

Explore lineage

Review live activity, lineage, policy decisions, alerts, and policy edits against the same endpoint store.

Open dashboard docs