What is KorClaw?
The runtime compliance layer that governs AI agent actions before they execute.
KorClaw sits between your applications and AI providers. When an agent attempts an action — calling an API, reading a file, invoking an MCP tool — KorClaw evaluates workspace policies, checks agent identity and allowlists, routes high-risk actions through approval, and records every decision in an audit log.
Core capabilities
- Policy engine — declarative rules with block, require_approval, alert, redact, log, and allow actions
- Agent registry — identity, environment, risk level, allowed tools and resources
- Runtime enforcement — governed HTTP proxy and MCP proxy execution
- Approval center — human-in-the-loop gates with replay support
- Audit logs — every evaluation and execution is recorded
- MCP governance — register servers, discover tools, call tools through KorClaw
KorClaw does not replace your AI stack. Your agents continue using OpenAI, Anthropic, Gemini, or custom providers. KorClaw governs what those agents are allowed to do.