Discover and explore top open-source AI tools and projects—updated daily.
radareorgRadare2 binary analysis server for AI agents
Top 95.5% on SourcePulse
Summary
This project provides an MCP (Machine Communication Protocol) server for radare2, enabling AI agents and IDEs to interact with radare2's powerful binary analysis capabilities. It bridges the gap between AI tools like Claude, VSCode Copilot, and Zed, and the deep analysis features of radare2, offering a seamless way to integrate reverse engineering workflows into AI-assisted development environments.
How It Works
The server is fully written in C, leveraging native radare2 APIs for deep integration. It operates via standard input/output (stdin/stdout) or as an HTTP server, allowing flexible connection to local or remote radare2 sessions via r2pipe. Key design choices include support for read-only modes, sandboxing, fine-grained tool configuration, and direct communication, facilitating secure and controlled AI-driven analysis.
Quick Start & Requirements
Installation is straightforward using r2pm -Uci r2mcp. For integration, clients like Claude Desktop, VS Code, and Zed require specific JSON configuration pointing to r2pm -r r2mcp as the command. Docker builds are also supported. Advanced HTTP features like session multiplexing (-X) and bearer authentication (-a) require radare2 ABI version 91 or newer. Building from source is possible via make install (Linux/macOS) or meson/ninja (Windows).
Highlighted Details
Maintenance & Community
The project is part of the radareorg ecosystem. Specific details on active contributors, sponsorships, or community channels (like Discord/Slack) are not detailed in this README snippet.
Licensing & Compatibility
The README does not explicitly state the license. However, radare2 projects are typically licensed under permissive licenses like LGPL or MIT, allowing for broad compatibility. Users should verify the specific license for commercial use.
Limitations & Caveats
Advanced HTTP features (session multiplexing, bearer authentication) are dependent on radare2 ABI version 91 or newer. The server requires a separate LLM agent to be functional. Some features like sandboxing and tool restrictions are optional and configurable.
3 weeks ago
Inactive