# Provenza

> Provenza is an open-source MCP server for security. Governed MCP gateway: approved tools only, each call policy-checked and recorded. 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://stickstock.lol
- **Repository:** https://github.com/kironovlaziz-del/provenza/tree/HEAD/sdk/python
- **MCP Registry name:** io.github.kironovlaziz-del/provenza
- **Categories:** security, AI models and LLM tooling
- **Listing updated:** 2026-11-15

## About

Provenza is open-source, self-hosted governance for AI agents, and this package lets an MCP client reach tools through it. Each call is checked against the agent's rights, organization policies, a tool registry, behaviour rules and a circuit breaker before it runs. Denied calls raise an error, calls needing a human decision are held until a person approves, and results are signed with the agent's own key and sealed into a tamper-evident audit log with receipts verifiable offline.

An admin issues a one-time enrollment token under Agents then New agent; keys are generated on the agent's machine and only the public key goes to the server. The Python SDK fails closed when the server is unreachable. Standalone mode writes a local signed, hash-chained log that stores hashes of arguments and results, never content. Optional keychain storage and post-quantum hybrid keys (Ed25519 with ML-DSA-65) are available.

It needs a self-hosted Provenza server and URL.

## Features

- **Policy checks:** Rights, policies, registry and behaviour checked before each call.
- **Human approval holds:** Calls needing approval wait, then run once after sign-off.
- **Signed audit log:** Results are signed and sealed in a tamper-evident log.
- **Fail-closed SDK:** Tools do not run if the server is unreachable.
- **Local identity keys:** Keys generated on the agent's machine.
- **Delegation chains:** Agents can delegate tasks with scoped tools and expiry.

## Good for

- Gate which tools an agent may call
- Require approval before a report is emailed
- Keep a verifiable record of agent actions
- Restrict a standalone agent to an allow list

## Install

### Claude Code (Terminal)

```sh
claude mcp add provenza --env PROVENZA_URL='<PROVENZA_URL>' --env PROVENZA_TOKEN='<PROVENZA_TOKEN>' -- uvx provenza
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "provenza": {
      "command": "uvx",
      "args": [
        "provenza"
      ],
      "env": {
        "PROVENZA_URL": "<PROVENZA_URL>",
        "PROVENZA_TOKEN": "<PROVENZA_TOKEN>"
      }
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "provenza": {
      "command": "uvx",
      "args": [
        "provenza"
      ],
      "env": {
        "PROVENZA_URL": "<PROVENZA_URL>",
        "PROVENZA_TOKEN": "<PROVENZA_TOKEN>"
      }
    }
  }
}
```

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

```json
{
  "servers": {
    "provenza": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "provenza"
      ],
      "env": {
        "PROVENZA_URL": "<PROVENZA_URL>",
        "PROVENZA_TOKEN": "<PROVENZA_TOKEN>"
      }
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "provenza": {
      "command": "uvx",
      "args": [
        "provenza"
      ],
      "env": {
        "PROVENZA_URL": "<PROVENZA_URL>",
        "PROVENZA_TOKEN": "<PROVENZA_TOKEN>"
      }
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "provenza": {
      "command": "uvx",
      "args": [
        "provenza"
      ],
      "env": {
        "PROVENZA_URL": "<PROVENZA_URL>",
        "PROVENZA_TOKEN": "<PROVENZA_TOKEN>"
      }
    }
  }
}
```

## FAQ

### Does it need a server?

Connected mode needs your self-hosted Provenza at PROVENZA_URL; standalone mode runs without one.

### What happens if the server goes down?

The SDK fails closed unless fail_open=True is set.

### Does the audit log store my data?

The local log stores only hashes of arguments and results.

### How is an agent enrolled?

With a one-time enrollment token from Agents then New agent.

### Is Provenza free?

Yes. Provenza is open source, so you can use and self-host it for free.

### How do I add Provenza to Claude Code?

Run this command in your terminal: claude mcp add provenza --env PROVENZA_URL='<PROVENZA_URL>' --env PROVENZA_TOKEN='<PROVENZA_TOKEN>' -- uvx provenza

### What are the best alternatives to Provenza?

The closest alternatives to Provenza are REA, Claude Code Ultimate Guide, Agent Roast, Midplane and Leash.

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [Claude Code Ultimate Guide](https://ai-powered.com/mcp/claude-code-ultimate-guide): Claude Code Ultimate Guide is an open-source MCP server for ai models and llm tooling.
- [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.
- [Eikino](https://ai-powered.com/mcp/eikino): Eikino is a freemium MCP server for security.

---
Source: https://ai-powered.com/mcp/provenza
