Introduction

KorClaw is the runtime compliance layer for AI agents. Register agents, define policies, require human approval for high-risk actions, enforce decisions at runtime, and maintain a complete audit trail — via REST API, SDK, or CLI.

What you can do today

• Register and manage AI agents with identity, environment, and allowlists
• Create declarative policies with rule-based evaluation
• Evaluate actions before execution and block or require approval
• Execute HTTP and MCP tool calls through a governed runtime proxy
• Register multiple MCP servers per workspace and call tools through KorClaw
• Create and decide approval requests
• Query audit logs and receive signed webhook events

Start here

1. Read What is KorClaw? and Architecture for the mental model
2. Follow Quick Start to evaluate your first policy
3. Explore the Runtime API reference for endpoint details
4. Install the TypeScript SDK or Python SDK for application integration