# Protect MCP

> Protect MCP is an open-source MCP server for security. Cedar policy gate and signed receipts for agent tool calls, local and fail-closed. It runs locally on your machine and is started by your AI client.

- **Kind:** MCP server
- **Pricing:** Open source
- **Platforms:** CLI, self-hosted
- **Repository:** https://github.com/scopeblind/scopeblind-gateway
- **MCP Registry name:** com.scopeblind/protect-mcp
- **Categories:** [security](https://ai-powered.com/categories/security), [developer tools](https://ai-powered.com/categories/developer-tools)
- **Listing updated:** 2026-11-18

## About

Placed between an MCP client and its tool server, or used through coding-agent hooks, this gateway evaluates each proposed call against a Cedar policy and signs the decision with Ed25519. In enforce mode any missing policy or engine error means deny. The MCP wrapper starts in shadow mode, logging what would be blocked, until you add the --enforce flag.

Receipts include the decision and a policy digest, can be chained, and are checked with a separate verifier tool; tampering with a signed deny turns up as an invalid signature. The README is frank about limits: it covers only calls routed through it, is not an OS sandbox, and a receipt does not prove an action ran or that nothing was omitted. Starter shell rules inspect command text and can be bypassed with alternate spellings.

Evaluation and verification run locally with no ScopeBlind account and no decision telemetry. Integrations include a Claude Code plugin, project hooks, an HTTP gateway and previews for Cursor and Codex.

## Features

- **Cedar policy evaluation:** Allow or deny each tool call from a reviewable policy.
- **Fail-closed mode:** Missing policies or engine errors result in deny.
- **Signed receipts:** Ed25519 receipts record the decision and policy digest.
- **Shadow-to-enforce workflow:** Observe calls first, then turn on blocking.
- **Several entry points:** MCP stdio wrapper, HTTP gateway, Claude Code plugin or hooks.

## Good for

- Block force-pushes and risky shell commands from a coding agent
- Keep an auditable record of what an agent was allowed to do
- Trial a policy in shadow mode before enforcing it
- Wrap an existing MCP server with access rules

## Install

### Claude Code (Terminal)

```sh
claude mcp add protect-mcp -- npx -y protect-mcp
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "protect-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "protect-mcp"
      ]
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "protect-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "protect-mcp"
      ]
    }
  }
}
```

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

```json
{
  "servers": {
    "protect-mcp": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "protect-mcp"
      ]
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "protect-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "protect-mcp"
      ]
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "protect-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "protect-mcp"
      ]
    }
  }
}
```

## FAQ

### Does it need an account?

No. Local evaluation and verification need no ScopeBlind account.

### Is it a sandbox?

No. It only covers calls routed through the gateway.

### What Node version does the demo use?

Node.js 22.

### Does a receipt prove the action ran?

No. The README says it does not establish execution or completeness.

### Is Protect MCP free?

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

### How do I add Protect MCP to Claude Code?

Run this command in your terminal: claude mcp add protect-mcp -- npx -y protect-mcp

### What are the best alternatives to Protect MCP?

The closest alternatives to Protect MCP are REA, Context7, Chrome DevTools MCP, Codebase Memory and GitHub.

## Where this comes from

- Description: written from the project's README.
- Install steps: from the official MCP Registry entry.

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [Context7](https://ai-powered.com/mcp/context7): Context7 is a freemium MCP server for developer tools.
- [Chrome DevTools MCP](https://ai-powered.com/mcp/chrome-devtools-mcp-2): Chrome DevTools MCP is an open-source MCP server for browser automation.
- [Codebase Memory](https://ai-powered.com/mcp/codebase-memory): Codebase Memory is an open-source MCP server for knowledge and memory.
- [GitHub](https://ai-powered.com/mcp/github-2): GitHub is an open-source MCP server for developer tools.
- [Serena MCP: the IDE for your agent](https://ai-powered.com/mcp/serena-mcp-the-ide-for-your-agent): Serena MCP: the IDE for your agent is an open-source MCP server for developer tools.
- [awesome-mcp-servers](https://ai-powered.com/mcp/awesome-mcp-servers-2): awesome-mcp-servers is an open-source MCP server for developer tools.
- [XERJ](https://ai-powered.com/mcp/xerj): XERJ is an open-source MCP server for knowledge and memory.

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