kakunin-mcp is a Model Context Protocol (MCP) server for AI agent identity and compliance scenarios, wrapping the Kakunin AI agent compliance API. It enables AI agents to independently perform permission verification, risk queries, and behavior audit logging within Claude, Cursor, or any MCP-compatible runtime.
Purpose The tool is used for AI agent compliance and identity management, helping agents confirm their own permissions before performing operations and continuously record auditable behavior trails.
Main capabilities
- verify_agent_scope: Before performing an action, check whether the agent is authorized to perform that action. It verifies the currently valid X.509 certificate, the permitted_actions permission scope, financial limits, and revocation status, and returns { allowed: true|false, reason, certificate_status, permitted_actions }.
- check_risk_score: Retrieve the agent's rolling 30-day risk score, risk level (low / medium / high), drift trend, and actionable recommendations. This operation requires no input.
- audit_log_append: Append a behavior event to the agent's tamper-proof audit log, returning a risk score and transaction ID. Events with a score ≥ 0.85 automatically trigger a certificate revocation check.
How to use Prerequisites: Node ≥ 18, and a Kakunin API key and Agent ID obtained from kakunin.ai/dashboard. Add the server to ~/Library/Application Support/Claude/claude_desktop_config.json (the specific configuration content is not provided in the source material). Environment variables: During development, you can use a key with the kak_test_... prefix; this key connects to the sandbox CA, has no cost, and provides 100 free certificates per day. Full documentation is available at docs.kakunin.ai.
Use cases Suitable for scenarios that require certificate verification, permission scope checks, risk score queries, and behavior audit logging for AI agents in Claude, Cursor, or other MCP-compatible runtimes.
Limitations The source material does not provide more information about fees, pricing, platform compatibility details, performance results, or specific configuration steps.
