Members-Only
Recent Talks & Demos are for members only
You must be an AI Tinkerers active member to view these talks and demos.
SecureClaw: Secure AI Agent Wrapper
Learn how SecureClaw builds a secure AI agent wrapper. This talk demonstrates a daily-driver workflow for agentic coding, preventing credential leakage and unauthorized access.
Carapace (a.k.a. SecureClaw) is a security layer for AI agents that prevents credential leakage and unauthorized access. It network-jails the agent in Docker with no DNS, routes all traffic through a MITM proxy with phone-based domain approval, and hides API credentials behind restrictive host-side proxies. The agent can read your email but never sees the API key, and can’t send without you.