Open-source Model Context Protocol servers that connect AI assistants to amateur radio services. Ask Claude, ChatGPT, Copilot, or Gemini about your QSOs, confirmations, and logbook data — no manual API wrangling required.

14 packages · 91 tools · one install command.

View the live demo → · Documentation →

curl -sL https://qso-graph.io/install.sh | bash

Creates ~/.qso-graph/ with an isolated Python environment, installs the base MCP servers (6 servers, 38 tools), and adds them to your PATH. Works on Linux and macOS. No root required.

After install, run qso-graph-config to manage servers, credentials, datasets, and MCP client configuration via an interactive TUI.

Advanced users can install directly via pip:

pip install qso-graph-config # Base — 6 servers, 45 tools
pip install "qso-graph-config[auth]"# + 4 logbook servers (26 tools)
pip install "qso-graph-config[ionis]"# + ionis-mcp propagation (12 tools)
pip install "qso-graph-config[full]"# Everything — 13 packages, 91 tools

Your credentials never leave the OS keyring. Every qso-graph server enforces non-negotiable security guarantees:

• Credentials stored in OS keyring only (macOS Keychain, Windows Credential Manager, Linux Secret Service) — never in config files, environment variables, or logs
• Credentials never appear in MCP tool results, error messages, or debug output — enforced by architecture, not policy
• No command injection surface — no subprocess, no shell=True, no eval
• All external connections HTTPS only
• Rate limiting on every API to prevent account bans
• Independent security audit required before every PyPI release

Full details: Security

# One-line install (Linux / macOS)
curl -sL https://qso-graph.io/install.sh | bash
# Launch the config manager
qso-graph-config
# Or install a single server directly
pip install solar-mcp

Each server works with any MCP client: Claude Desktop, Claude Code, ChatGPT, Cursor, VS Code / GitHub Copilot, Windsurf, Gemini CLI, Goose, and Codex CLI — or use qsp-mcp to relay tools to any local LLM.

install.sh (bootstrap) qso-graph-config (manager)
└── ~/.qso-graph/ ├── Install / upgrade servers
├── venv/ ├── Credential setup (qso-auth)
├── bin/ (PATH) ├── Dataset downloads (ionis-mcp)
└── etc/state.json └── MCP client config generation
qso-graph-auth (identity) MCP Servers (qso-graph)
├── PersonaManager ──> eqsl-mcp, qrz-mcp, lotw-mcp, hamqth-mcp
├── OS keyring credentials Each server = 1 pip install
└── qso-auth CLI Each server = 4-8 MCP tools
adif-mcp (ADIF spec) Public Servers
└── 8 spec tools ──> solar, pota, sota, iota, wspr (no auth)
qsp-mcp (tool relay) Local LLM Inference
└── Stateless pipe ──> llama.cpp, Ollama, vLLM, SGLang
MCP tools → OpenAI Any model with function calling
tools format Zero cloud dependency

Do NOT open public GitHub issues for security vulnerabilities. Email: ki7mt@yahoo.com with subject [SECURITY] qso-graph vulnerability report.

GPL-3.0-or-later