SafeClaw in 60 Seconds: Quickstart Guide

SafeClaw is action-level gating for AI agents. Deny-by-default. Sub-millisecond. Zero dependencies. Here is how to go from nothing to protected in 5 steps.

Step 1: Install

npx @authensor/safeclaw

That is the entire install. Zero third-party dependencies. MIT license. 446 tests.

Step 2: Get Your Free Key

Go to safeclaw.onrender.com. Create an account. No credit card. You get a 7-day renewable API key instantly.

Step 3: Connect Your Agent

The setup wizard at safeclaw.onrender.com shows integration steps for your framework: Claude, OpenAI, LangChain, CrewAI, AutoGen, MCP, Cursor, Copilot, or Windsurf.

Step 4: Run in Simulation Mode

Start with simulation mode. SafeClaw evaluates every agent action — file_write, file_read, shell_exec, network — against your policy and logs the decision, without blocking anything. Review the results in the dashboard to see what your agent is actually doing.

Step 5: Enforce

When you are satisfied with the policy, switch to enforcement mode. Denied actions are now blocked before execution. Every decision is logged to a tamper-proof SHA-256 hash chain audit trail.

What You Now Have

• Deny-by-default policy on every agent action
• Sub-millisecond policy evaluation (no slowdown)
• Tamper-proof audit trail of every action
• Full visibility via browser dashboard
• Control plane sees only metadata — never your keys or data

Next Steps

• Customize your policy rules per action type
• Add human-in-the-loop review for high-risk actions
• Export audit data for compliance (SOC 2, HIPAA, GDPR)
• Read Your First SafeClaw Policy in 5 Minutes for detailed policy configuration