Gensee Crate CLI
A control layer for autonomous agents: action policies, review queue, recovery points and config audit for Claude Code and Codex.
- 125 ★ on the repository
- Last commit
Extended access
Why this level
- Hooks observe agent actions and write them to a local store.
- Installation uses a `curl ... | bash` script from the `main` branch.
Install
- Works in
Manual install
curl -fsSL https://raw.githubusercontent.com/GenseeAI/gensee-crate/main/scripts/install_oss.sh | bashInstalls the `gensee` CLI and prepares the local store; read the script before running it.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Hooks observe agent actions and write them to a local store.
- Installation uses a `curl ... | bash` script from the `main` branch.
This is third-party code. Review the repository files before installing.
What it does
Gensee Crate puts a local control layer around AI agents. The `gensee` CLI attaches hooks to Codex and Claude Code, records events in a local store and applies policy: allow, ask or block a supported action and protect sensitive paths. The review queue groups work by its originating request and shows scope drift, blocked actions and failed verification. Recovery points are created before risky changes in a git workspace. The configuration audit inspects instructions, skills, MCP servers, hooks and permissions without running the agent. A macOS app is available, and the Team edition provides isolated workspace forks on a self-hosted Linux host.
Who it is for. Developers on macOS and Linux who run Codex or Claude Code with broad permissions and want to see and limit agent actions.
Good fit when
- An agent ran in the background and you want to see which files it touched and whether it left the request scope.
- Before adding a third-party skill or MCP server you audit the agent configuration for risky spots.
- The same reads and commands repeat, and you approve them once per session or project.
- A team runs agents on a dedicated Linux host with isolated workspace copies.
Not a fit when
- You need hard isolation: actions are blocked only where the integration and environment support it.
- Work happens outside a git repository: recovery points cover supported changes in a git workspace.
- You need a mature production product: the project is in alpha.
Example request
Run Claude Code through Gensee and show which actions needed review.Limitations
The project is in alpha. Policy enforcement depends on the integration and environment: supported tool actions can be blocked, while commands inside the Claude Cowork guest VM and cloud execution stay outside coverage. The Personal app runs on macOS, and the Team edition targets a self-hosted Linux host. Recovery points cover supported changes in a git workspace. Policy and events live in a local directory set by `GENSEE_HOME`, `~/.gensee` by default.
How to disable. Remove the hooks through integration settings or by hand from the Codex and Claude Code configuration, then delete the `gensee` binary and the `GENSEE_HOME` directory.
FAQ
Which agents are supported?
Codex, Claude Code, Cursor, GitHub Copilot, Antigravity and Omnigent. A visibility pilot covers Claude Cowork on macOS.
How do I connect Claude Code?
Run `gensee setup claude-code --gensee-home "$GENSEE_HOME"`. `gensee run -- claude` starts a managed session.
How do I see results?
`gensee status --json` shows state, `gensee timeline --latest` the latest events, and `gensee audit codex` inspects the Codex configuration.
Is the app required?
No. The CLI works in a terminal, in automation and on Linux; on macOS the app is the recommended interface.
What does the Review Queue do?
It groups background activity with its originating request and surfaces only what needs a decision: scope drift, blocks, approval requests and failed verification.
Discussion
No questions or reviews yet.
Related
CLI toolsThe safehouse command runs Claude Code, Codex and other agents in a macOS sandbox: files are closed by default and you open folders with flags
A policy gateway for AI agents: it checks tool calls before they run, holds risky ones for approval and writes an audit trail
A local dashboard for Claude Code, Codex CLI and OpenCode: what the agent loads, its token cost and an MCP server security scan
A local guard for coding agents: blocks secret writes and force pushes to main, warns about risky commands.
Container-free sandbox for AI agents on Linux and macOS: files, network and syscalls stay closed until you open them explicitly
Claude Code, Codex, OpenCode and pi run in a hardened container: project files only, allowlisted network, keys held by a proxy.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Gensee Crate (this card) | no | 125 | |
| HOL Guard | yes | 1,072 | HOL Guard checks commands and installs before execution; Gensee Crate adds a review queue, recovery points and a configuration audit |
| CC Safety Net | no | 1,932 | cc-safety-net blocks destructive commands; Gensee Crate groups activity by request and runs an allow, ask, block policy |
README in short
The project splits into Personal for a laptop and Team for remote Linux environments. Both keep policy, action evidence and decisions in one place and keep routine activity out of the review queue.