MCP.so
登录
服务器

IDA Pro MCP

@mrexodia

AI-powered reverse engineering assistant that bridges IDA Pro with language models through MCP.

概览

What is IDA Pro MCP?

IDA Pro MCP is a simple MCP (Model Context Protocol) server that connects AI coding assistants to IDA Pro for reverse engineering, enabling "vibe reversing" by allowing LLMs to interact with a binary database through MCP tools. It is intended for users who want to analyze binaries using AI assistants alongside IDA Pro.

How to use IDA Pro MCP?

Install prerequisites (Python 3.11+, IDA Pro 8.3+, and a supported MCP client). For headless use with Claude Code, install via the Claude marketplace plugin; for GUI use, install the pip package and run ida-pro-mcp --install. Then start the server with uv run idalib-mcp --stdio or over HTTP/SSE, and configure your MCP client with the server's JSON configuration (obtainable via ida-pro-mcp --config).

Key features of IDA Pro MCP

  • Connects LLMs to IDA Pro via MCP tools.
  • Headless operation via idalib‑mcp supervisor.
  • GUI plugin for interactive use (will be deprecated).
  • Persistent, self‑managing database workers.
  • Supports both stdio and HTTP/SSE transports.

Use cases of IDA Pro MCP

  • Analyze a crackme with an LLM that adds comments and renames variables.
  • Perform automated malware analysis while documenting findings in markdown.
  • Open multiple databases simultaneously and switch between sessions.

FAQ from IDA Pro MCP

What are the prerequisites?

Python 3.11 or higher, IDA Pro 8.3 or higher (not IDA Free), and a supported MCP client.

Does IDA Pro MCP work headlessly?

Yes. After installing idalib, you can run uv run idalib-mcp --stdio or over HTTP/SSE to start a headless server.

How do I configure the MCP client?

Run ida-pro-mcp --config to get the JSON configuration block for your MCP client.

What transports are supported?

Stdio (for clients like Claude Code) and HTTP/SSE (bind to an address and port, e.g. --transport http://127.0.0.1:8744/sse).

What happens if I open a binary that is already loaded?

The headless worker adopts the existing session transparently; there is no separate "shared" mode to enable.

标签

来自「其他」的更多内容