MCP server for GNU Radio โ programmatic/LLM-driven generation and control of GNURadio flow-graphs (.grc).
https://github.com/yoelbassin/gr-mcpFinally, an MCP server that speaks GNU Radio. If you've spent hours clicking and dragging blocks in GNU Radio Companion, manually wiring up flowgraphs for every SDR experiment, you know the pain. GR-MCP changes that by bringing your AI assistant directly into your SDR workflow.
You're an experienced RF engineer or SDR researcher. You know exactly what signal processing chain you need, but translating that mental model into a GNU Radio flowgraph means opening GRC, hunting through block categories, configuring parameters, and connecting everything manually. Every. Single. Time.
GR-MCP eliminates this friction. Describe your SDR application to Claude, ChatGPT, or any MCP-compatible AI, and watch it generate the .grc file directly. No more point-and-click interface when you just want to prototype a new demodulator or test a filter design.
Rapid Prototyping: "Create a narrowband FM receiver for 146.52 MHz with a 12.5 kHz deviation" - and get a working flowgraph in seconds, not minutes.
Educational Content: Generate multiple variations of the same concept for teaching. Create basic AM demodulator examples, then complex ones, all programmatically.
Batch Processing: Need 20 similar flowgraphs for different frequencies or parameters? Script it instead of building each one manually.
Documentation and Reproducibility: Your AI assistant can generate flowgraphs with consistent naming conventions and proper documentation embedded in the blocks.
Integration Testing: Automatically generate test flowgraphs for validating custom GNU Radio blocks or comparing different implementation approaches.
This isn't a toy - it's designed for serious GNU Radio development. The server integrates with your existing GNU Radio installation (v3.10.12.0+ tested), handles the complex XML structure of .grc files correctly, and includes example flowgraphs to get you started.
The MCP interface means it works with any compatible AI client - Claude Desktop, Cursor, or your own custom automation scripts. You're not locked into a specific chat interface or proprietary system.
git clone https://github.com/yoelbassin/gr-mcp
cd gr-mcp
uv venv --system-site-packages # Respects your GNU Radio installation
Add to your MCP client configuration and start generating flowgraphs through natural language. The --system-site-packages flag ensures the server can access your existing GNU Radio Python modules.
Check the misc/ directory for example .grc files that demonstrate the server's capabilities.
This is active development, but the core functionality works today. The maintainer is responsive to feedback and the codebase is clean Python with proper testing via pytest. If you're doing regular GNU Radio work and want to see where AI-assisted SDR development is heading, this is worth experimenting with now.
The GNU Radio community has been relatively slow to adopt modern development tools. GR-MCP represents exactly the kind of automation that can make SDR development more productive and accessible.