2026-02-18 · John Kearney

Using SafeClaw with Claude Code

Clawdbot has leaked over 1.5 million API keys in under a month. Claude Code is powerful. It reads your files, runs shell commands, writes code, and makes network requests on your behalf. That is what makes it useful. It is also what makes it risky if you are working with sensitive files or production systems.

SafeClaw lets you keep all that power while adding per-action control.

Setup

No config files. No CLI expertise. Set it up in your browser.

npx @authensor/safeclaw

Browser opens with the setup wizard. It walks you through creating your first policy.

Example Policy for Claude Code

Here is a practical starting point:

File access:


Shell commands:

Network:

What Happens When Something Gets Blocked

The browser dashboard shows every blocked action with full context: what the agent tried to do, which policy rule blocked it, and a timestamp. You can approve the action retroactively or update your policy to allow similar actions in the future.

Simulation mode lets you test a policy before enforcing it. Run Claude Code normally, and SafeClaw will log what would have been blocked without actually blocking anything.

How Data is Handled

The client is 100% open source. The Authensor control plane is hosted and only sees action metadata, never your keys or data. Free tier with renewable 7-day keys. No credit card.

The Point

You do not have to choose between giving Claude Code full access or restricting it so much it cannot work. SafeClaw gives you the middle ground: full capability with per-action control. Accessible to non-developers. Set it up in your browser.

Install: npx @authensor/safeclaw
GitHub: https://github.com/AUTHENSOR/SafeClaw

Try SafeClaw

Action-level gating for AI agents. Set it up in your browser in 60 seconds.

$ npx @authensor/safeclaw