SDR MCP Server
sandraschi/sdr-mcp
Escuta ondas de rádio com um stick USB, controlado por Claude ou dashboard web
Ver no GitHubServidor MCP que transforma um dongle RTL-SDR USB em um receptor de rádio controlado por IA ou dashboard web. Fornece espectro em tempo real, waterfall, demodulação FM, banco de estações e integração com GNU Radio via sidecar Docker.
Instalação
npx convez-ai sandraschi/sdr-mcp1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Monitoramento de espectro RF em tempo real via Claude
- Recepção e demodulação FM automática com IA
- Pesquisa de engenharia de rádio e SDR educacional
- Integração de processamento de sinais em aplicações Claude
Veredito

Ferramenta especializada de alta complexidade técnica que abre capacidades de processamento de sinais RF para Claude. Requer hardware específico e conhecimento em rádio/engenharia, mas oferece controle pioneiro de SDR via MCP com dashboard robusto.
README
A tradução para o português deste README ainda está na fila.
SDR MCP Server
Listen to radio waves with a USB stick — controlled by AI or a web dashboard.
Software Defined Radio (SDR) turns RF into data your PC can plot and play. This repo is a Model Context Protocol server for RTL-SDR: live spectrum, waterfall, FM audio, station databases, and a GNU Radio demod sidecar. Works with Cursor, Claude Desktop, or the built-in React dashboard at http://127.0.0.1:10890/.
Quick Start
git clone https://github.com/sandraschi/sdr-mcp
cd sdr-mcp
just
This opens an interactive dashboard showing all available commands. Run just bootstrap to install dependencies, then just serve or just dev to start.
Manual Setup
If you don't have just installed:
pip install sdr-mcp
sdr-mcp check # needs RTL-SDR + WinUSB on Windows, or use mock mode
sdr-mcp serve # STDIO for Claude Desktop
Windows + real dongle: replace the stick's DVB/TV USB driver with WinUSB via Zadig — see INSTALL.md. You are not replacing the hardware; only the driver Windows uses to talk to USB.
No dongle: mock mode activates automatically (SDR_MCP_MOCK=auto). FFT/waterfall still work.
See MOCK_SDR.md.
For the web dashboard:
cd web_sota
npm install
npm run dev
Or double-click web_sota\start.bat.
Documentation
| Document | What it covers |
|---|---|
| INSTALL.md | Full setup, drivers, configuration |
| ARCHITECTURE.md | System design, data flow, transport |
| MCP_SERVER.md | Portmanteau MCP tools with examples |
| SDR_TECHNOLOGY.md | Radio basics for beginners |
| RTL_SDR_V4.md | Recommended hardware specs |
| GNURADIO.md | GNU Radio sidecar (Docker + rtl_tcp) |
| HACKRF.md | HackRF TX licensing + hardware buying guide |
| MOCK_SDR.md | Synthetic IQ demo mode (no dongle) |
| OSCILLOSCOPE_MCP.md | USB oscilloscope MCP feasibility |
New to SDR? Open the dashboard → Help (horizontal tabs) or read SDR_TECHNOLOGY.md.
GitHub topics
sdr · software-defined-radio · rtl-sdr · mcp · mcp-server · fastmcp · gnuradio · spectrum-analyzer · waterfall · websocket · python · react · docker
See .github/REPOSITORY.md for gh repo edit commands.
Features
Hardware Control
- Auto-detect RTL-SDR devices, initialize and configure
- Mock IQ mode when no dongle — FFT/waterfall/WebSocket still work (MOCK_SDR.md)
- Set frequency (24 MHz — 1.766 GHz), gain (auto or manual)
- Real-time IQ sample capture and spectrum processing
Spectrum Analysis
- 2048-point FFT with Hamming window
- Peak detection and signal strength analysis
- Waterfall history (100 lines) for time-varying signals
GNU Radio Demod (sidecar)
- FM demod via Dockerized GNU Radio + rtl_tcp
- MCP tool
sdr_gnuradiofor start/stop/status - HackRF path documented for future SoapySDR integration
Frequency Database
- 11 pre-loaded stations across LW/MW/SW/VHF bands
- Program schedules with current-playing info
- Online search via radio-browser.info (25k+ stations)
WebSocket Streaming
-
Real-time spectrum broadcast to web clients
-
Remote frequency/gain control via WebSocket commands
-
Canvas-based spectrum and waterfall visualizations
-
Web Dashboard — noob-friendly hero, tabbed Help, spectrum + waterfall + FM audio
Hardware
Platform: Developed and tested on Windows (WinUSB via Zadig, fleet launchers, MCPB bundle). Core Python code may run on Linux with librtlsdr, but the dashboard launch scripts and MCPB manifest target Windows only.
Recommended: RTL-SDR Blog v4 (~$35)
- 24 MHz — 1.766 GHz continuous coverage
- 0.5 ppm TCXO for frequency stability
- SMA connector, aluminum enclosure, bias tee
Any RTL2832U-based SDR with R820T2 tuner works. See RTL_SDR_V4.md for full specs.
Project Structure
sdr-mcp/
├── README.md # This file
├── docs/ # Documentation
│ ├── INSTALL.md # Setup guide
│ ├── ARCHITECTURE.md # System design
│ ├── MCP_SERVER.md # Tool reference
│ ├── SDR_TECHNOLOGY.md # Radio primer
│ └── RTL_SDR_V4.md # Hardware specs
├── pyproject.toml # Python package config
├── justfile # Lint, fix, security recipes
├── start.ps1 # Launch backend + webapp
├── src/sdr_mcp/ # Python backend
│ ├── server.py # FastMCP server, 17 tools
│ ├── capture.py # RTL-SDR hardware interface
│ ├── processor.py # FFT / spectrum processing
│ ├── frequency_db.py # Station database
│ ├── online_db.py # radio-browser.info API
│ ├── websocket_server.py # Real-time WebSocket stream
│ ├── transport.py # STDIO / HTTP transport
│ └── cli.py # Command-line interface
├── web_sota/ # React/TypeScript webapp
│ └── src/
│ ├── pages/ # Spectrum, Waterfall, Stations, etc.
│ └── components/ # Layout, UI components
└── tests/ # Pytest test suite
Tech Stack
| Layer | Technology |
|---|---|
| Protocol | FastMCP 3.4, MCP 2.14+ |
| Backend | Python 3.12, asyncio |
| Hardware | pyrtlsdr, RtlSdr |
| Signal | numpy, scipy (FFT) |
| Streaming | websockets (RFC 6455) |
| Frontend | React 19, TypeScript, Vite |
| UI | Tailwind CSS, Radix UI, Lucide icons |
| Standards | Fleet SOTA, ruff, Biome, just |
License
MIT