# kakunin [Health: Active]

**Category:** 📊 Monitoring  
**Repository:** https://github.com/nqzai/kakunin-mcp  
**GitHub Stars:** 1  
**npm Downloads (last month):** 213  
**Views:** 1  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/kakunin-2

## Description
X.509 identity, risk scoring, and audit logging for AI agents. MiCA + EU AI Act compliant.

## Tools
Capabilities this server exposes over MCP:

- **verify_agent_scope** — Check whether this agent is authorised to perform an action before executing it. Verifies the active X.509 certificate, permitted_actions scope, financial limits, and revocation status.
- **check_risk_score** — Retrieve the agent's rolling 30-day risk score, band (`low`/`medium`/`high`), drift trend, and actionable guidance. No input required.
- **audit_log_append** — Append a behavioral event to the agent's immutable audit log. Returns risk score + transaction ID. Events scoring ≥ 0.85 auto-trigger a certificate revocation check.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "kakunin": {
    "command": "npx",
    "args": ["-y","@kakunin/mcp"],
    "env": {
      "KAKUNIN_API_KEY": "",
      "KAKUNIN_AGENT_ID": ""
    }
  }
}
```

**Requires environment variables:** `KAKUNIN_API_KEY`, `KAKUNIN_AGENT_ID` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What kakunin MCP server does

The kakunin MCP server gives an AI agent access to three Kakunin compliance operations. It can check whether a proposed action is authorized, retrieve the agent’s current risk information, and record behavioral events. The service acts for one configured Kakunin agent identified by `KAKUNIN_AGENT_ID`.

Authorization checks use the active X.509 certificate, the agent’s permitted actions, configured financial limits, and certificate revocation status. This makes `verify_agent_scope` suitable for checking a transaction or other operation before the agent carries it out. The tool returns an allow or deny result, an explanation, certificate status, and permitted actions.

The project description positions Kakunin around AI agent identity, risk scoring, audit logging, and MiCA plus EU AI Act compliance. The repository specifically documents connections to the Kakunin compliance API rather than to a transaction venue or other external system.

## How it works

The server communicates with Kakunin using credentials supplied through environment variables. An MCP client starts the server with Node.js and exposes its tools to the agent. For an authorization request, the agent supplies an action description, a venue, and an amount in U.S. dollars. The Kakunin API evaluates those details against the configured identity and permissions.

`check_risk_score` needs no input. It returns a rolling 30-day score, a `low`, `medium`, or `high` band, a drift trend, and guidance. `audit_log_append` accepts an action type and detail object, then returns a risk score and transaction ID. Events with a score of at least 0.85 automatically cause a certificate revocation check.

## Setup and configuration

Install requirements are Node.js 18 or newer, a Kakunin API key, and an agent ID obtained from the Kakunin dashboard. The documented launch command is:

```bash
npx -y @kakunin/mcp
```

Configure the command as an MCP server in the client’s settings and provide `KAKUNIN_API_KEY` and `KAKUNIN_AGENT_ID`. `KAKUNIN_BASE_URL` can override the default API base URL, which is `https://kakunin.ai`.

For development, use a `kak_test_...` key. The sandbox uses a sandbox certificate authority, does not incur cost, and includes 100 certificates per day. Production keys use the `kak_live_...` format.

## Tools and capabilities

- `verify_agent_scope`: checks certificate validity, action permissions, financial limits, and revocation status for a proposed action.
- `check_risk_score`: returns the rolling 30-day risk score, risk band, drift direction, and guidance without requiring parameters.
- `audit_log_append`: records a behavioral event and returns its risk score and transaction ID.
- Automatic revocation checking occurs for logged events scoring 0.85 or higher.

## Limitations and notes

The kakunin MCP server requires a Kakunin API key and a specific agent ID; it is not a standalone identity or scoring system. The documented inputs and outputs cover authorization checks, risk retrieval, and audit events, but do not describe tools for issuing certificates, changing permissions, or managing agents.

Use sandbox credentials for development. The README identifies Claude Desktop and Cursor as example clients and states that the server can run in any MCP-compatible runtime.

_Full upstream README: https://allmcps.com/mcp/kakunin-2/readme_