# Frame MCP

> Frame MCP is an open-source MCP server for security. Let a coding agent scan, verify and fix vulnerabilities with the Frame scanner. 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/lambdasec/frame-mcp
- **MCP Registry name:** io.github.lambdasec/frame-mcp
- **Categories:** [security](https://ai-powered.com/categories/security), [developer tools](https://ai-powered.com/categories/developer-tools)
- **Listing updated:** 2026-11-18

## About

Frame pairs taint analysis with separation logic verification, so each reported data-flow path (SQL injection, command injection, path traversal and similar) is checked by a solver. This package wraps the frame command-line program, launching it as a subprocess with an explicit argument list and never through a shell.

Beyond scan, it exposes the verify, fix and exploit steps, framework-pack checks, a solve tool for separation logic entailments, and a set of engine_* queries covering handlers, callers, callees, guards and traces. Those engine queries start a Frame query server (Frame 1.1.5 or later) on first use and answer later calls from memory, rebuilding when a source file changes. A security-review prompt chains scan, verify and an offer to fix.

Plain scanning needs no model. The AI options (scan with ai=true, fix, exploit) call an OpenAI-compatible endpoint configured with FRAME_LLM_BASE_URL, FRAME_LLM_API_KEY and FRAME_LLM_MODEL. The exploit tool sends live traffic, so the README says to run it only against systems you own or are authorized to test.

## Features

- **Taint plus solver checks:** Findings are verified with separation logic rather than pattern matches alone.
- **Fix and re-verify:** Generate fixes as diffs and re-scan to confirm they hold.
- **Engine queries:** Inspect handlers, callers, callees, guards and traces in a project.
- **Baselines:** Record current findings as a baseline file.
- **Optional AI layer:** Add LLM detection and triage through an OpenAI-compatible endpoint.
- **Many clients:** Install steps for Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI and Zed.

## Good for

- Scan a repository for injection flaws from an agent session
- Confirm that a patch actually closed a finding
- Check whether a handler is guarded on every path
- Hunt for missing authorization checks

## Install

### Claude Code (Terminal)

```sh
claude mcp add frame-mcp --env FRAME_LLM_API_KEY='<FRAME_LLM_API_KEY>' -- uvx frame-mcp
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "frame-mcp": {
      "command": "uvx",
      "args": [
        "frame-mcp"
      ],
      "env": {
        "FRAME_LLM_API_KEY": "<FRAME_LLM_API_KEY>"
      }
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "frame-mcp": {
      "command": "uvx",
      "args": [
        "frame-mcp"
      ],
      "env": {
        "FRAME_LLM_API_KEY": "<FRAME_LLM_API_KEY>"
      }
    }
  }
}
```

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

```json
{
  "servers": {
    "frame-mcp": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "frame-mcp"
      ],
      "env": {
        "FRAME_LLM_API_KEY": "<FRAME_LLM_API_KEY>"
      }
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "frame-mcp": {
      "command": "uvx",
      "args": [
        "frame-mcp"
      ],
      "env": {
        "FRAME_LLM_API_KEY": "<FRAME_LLM_API_KEY>"
      }
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "frame-mcp": {
      "command": "uvx",
      "args": [
        "frame-mcp"
      ],
      "env": {
        "FRAME_LLM_API_KEY": "<FRAME_LLM_API_KEY>"
      }
    }
  }
}
```

## Tools

- `scan`: Scan a file or directory for vulnerabilities
- `write_baseline`: Record findings as a baseline file
- `verify`: Re-scan to check findings are fixed
- `fix`: Generate and verify fixes
- `exploit`: Drive an agent to exploit an authorized live target
- `packs_list`: List framework packs
- `packs_check`: Validate framework pack files
- `solve`: Check a separation logic entailment
- `engine_handlers`: List entry points the engine finds
- `engine_handler_facts`: Facts about one handler or function
- `engine_siblings`: Handlers of the same class or file
- `engine_trace`: Data flows in a function
- `engine_callers`: Callers of a function
- `engine_callees`: Callees of a function
- `engine_guards`: Conditions that hold on paths to a line
- `engine_check`: Decide a claim about a line

## FAQ

### How do I install it?

Run uvx frame-mcp, or add it to your client with the stanza shown for Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI or Zed.

### Does it need an LLM?

Not for plain scans. The AI scan, fix and exploit tools call an OpenAI-compatible endpoint.

### Is the exploit tool safe to run anywhere?

No. Use it only on systems you own or are authorized to test.

### What licence?

Apache-2.0.

### Is Frame MCP free?

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

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

Run this command in your terminal: claude mcp add frame-mcp --env FRAME_LLM_API_KEY='<FRAME_LLM_API_KEY>' -- uvx frame-mcp

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

The closest alternatives to Frame 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/frame-mcp
