Every coding agent session ends one of two ways: you babysit permission prompts all day, or you run --dangerously-skip-permissions and pray. Clawk’s bet: those flags aren’t reckless if the agent isn’t on your machine. Show HN agreed — 225 points, 157 comments.
What it is
Clawk is an open-source CLI that boots a lightweight VM per project — Apple’s Virtualization.framework on macOS, Firecracker on Linux — mounts your repo inside, and lets Claude Code, Codex, or any shell agent run fully autonomous behind the wall. cd into a repo, type clawk, done. No Docker daemon, no config. Destroy the VM anytime; agent conversations and state survive.
The details that matter
Outbound traffic passes a DNS-aware allow-list — npm, PyPI, GitHub pre-approved, everything else blocked, so a prompt-injected agent can’t exfiltrate to random hosts. SSH-agent forwarding keeps commit signing working. Idle VMs are reclaimed automatically, and each sandbox is a copy-on-write clone, so disk cost is near zero.
“Full autonomy vs. safety” is the hottest fight in coding agents right now, and sandboxing is the consensus answer. Clawk just shrank it to one word. Apache 2.0, macOS 14+ on Apple silicon; Linux support is still experimental.
You Might Also Like
- 27k Github Stars in Weeks Learn Claude Code by Shareai lab Breaks Down ai Coding Agents Into 12 Lessons
- Claude hud hit 5 3k Github Stars Because Developers Were Flying Blind With Claude Code
- 27 Agents 109 Skills 88k Github Stars is Everything Claude Code Genius or Over Engineering
- From Claude Flow to Ruflo 22k Stars 5900 Commits and the Multi Agent Swarm Taking Over Claude Code
- Oh my Claudecode Turns Claude Code Into a 32 Agent dev Team and 11 8k Stars Agree
Top AI Product 
Leave a comment