# AMR (Agent Mandate Registry)

> AMR (Agent Mandate Registry) is an open-source MCP server for security. Registry that records who authorised an AI agent, within what limits and until when. It runs locally on your machine and is started by your AI client.

- **Kind:** MCP server
- **Pricing:** Open source
- **Platforms:** CLI, self-hosted
- **Website:** https://mandatia.eu
- **Repository:** https://github.com/jc7333/amr
- **MCP Registry name:** io.github.JC7333/mandatia
- **Categories:** security, legal and government
- **Listing updated:** 2026-11-14

## About

AMR starts from a gap: an organisation decides what an AI tool may do, but nothing enforces that once the agent acts alone. Here the principal, a person or a body, writes a mandate listing allowed actions, forbidden actions, spending caps and an end date, then signs it. Before each action the agent asks the registry for a token, and outside the mandate it gets none.

The website's demo is a medical-secretary assistant: booking an appointment gets a five-minute token, an 80 euro order gets a 30-second token, a 2,000 euro order is refused because the mandate caps orders at 100 euros, and prescriptions and patient messages are forbidden. A receiving third party can verify tokens in seconds. A hash-chained journal records the trail.

Run it locally with uvx mandatia; it uses SQLite and an Ed25519 key and makes no network calls by default. Tool profiles (principal, agent, receiver) keep an agent from authorising itself. The README frames it as support for EU AI Act Article 26 record-keeping, not as compliance on its own.

## Features

- **Signed mandates:** Principals set allowed and forbidden actions, caps and an end date.
- **Action tokens:** Short-lived signed tokens issued only inside the mandate.
- **Hash-chained journal:** Tamper-evident audit trail of authorisations.
- **Tool profiles:** Separate tool sets for principal, agent and receiver.
- **Third-party check:** Receivers verify a token without trusting the agent.
- **Local by default:** Runs on your machine with a SQLite file.

## Good for

- Limit a procurement agent to a spend cap
- Prove who authorised an agent to act
- Let a counterparty verify an agent's authority
- Keep records for AI Act oversight duties
- Create sub-mandates under an existing mandate

## Install

### Claude Code (Terminal)

```sh
claude mcp add mandatia -- uvx mandatia
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "mandatia": {
      "command": "uvx",
      "args": [
        "mandatia"
      ]
    }
  }
}
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "mandatia": {
      "command": "uvx",
      "args": [
        "mandatia"
      ]
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "mandatia": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "mandatia"
      ]
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "mandatia": {
      "command": "uvx",
      "args": [
        "mandatia"
      ]
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "mandatia": {
      "command": "uvx",
      "args": [
        "mandatia"
      ]
    }
  }
}
```

## Tools

- `create_mandate`: Register who authorises what for an agent.
- `register_principal_key`: Register a principal's key.
- `revoke_mandate`: Revoke a mandate.
- `contest_action`: Contest an action.
- `get_proof`: Get proof of a mandate.
- `export_insurer_dossier`: Export a dossier for an insurer.
- `verify_mandate`: Check whether an agent is authorised for an action.
- `issue_action_token`: Issue a signed action token.
- `log_action`: Log an action.
- `issue_mandate_attestation`: Issue a mandate attestation.
- `get_registry_public_key`: Get the registry's public key.
- `check_action_token`: Verify an action token as a receiver.

## FAQ

### Does it make a system compliant?

No. The README says it is a technical tool that does not by itself make a system compliant.

### How do I install it?

Run uvx mandatia, or claude mcp add amr -- uvx mandatia.

### Can an agent grant itself a mandate?

Not with the agent profile: root mandates are refused without a parent.

### Does it call out to the network?

Not by default; issue reporting and did:web key checks are opt-in.

### Is AMR (Agent Mandate Registry) free?

Yes. AMR (Agent Mandate Registry) is open source, so you can use and self-host it for free.

### How do I add AMR (Agent Mandate Registry) to Claude Code?

Run this command in your terminal: claude mcp add mandatia -- uvx mandatia

### What are the best alternatives to AMR (Agent Mandate Registry)?

The closest alternatives to AMR (Agent Mandate Registry) are REA, Agent Roast, Midplane, Leash and Tinplate Edge AI Safety, Security & EU Compliance.

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [Agent Roast](https://ai-powered.com/mcp/agent-roast): Agent Roast is an open-source MCP server for developer tools.
- [Midplane](https://ai-powered.com/mcp/midplane): Midplane is an open-source MCP server for databases.
- [Leash](https://ai-powered.com/mcp/leash): Leash is a free MCP server for security.
- [Tinplate Edge AI Safety, Security & EU Compliance](https://ai-powered.com/mcp/tinplate-edge-ai-safety-security-and-eu-compliance): Tinplate Edge AI Safety, Security & EU Compliance is a freemium MCP server for security.
- [SecureCloud](https://ai-powered.com/mcp/securecloud): SecureCloud is a free MCP server for security.
- [NY LLC Publication (FileMyPublication)](https://ai-powered.com/mcp/ny-llc-publication-filemypublication): NY LLC Publication (FileMyPublication) is a free MCP server for legal and government.
- [Eikino](https://ai-powered.com/mcp/eikino): Eikino is a freemium MCP server for security.

---
Source: https://ai-powered.com/mcp/amr-agent-mandate-registry
