☀️ solar-mcp

US solar data, agent-accessible.

Open-source MCP servers that bring US open solar data to ChatGPT, Claude, and anything else that speaks MCP.

What you can ask

Compare annual production for an 8 kW system in Mesa, AZ at 10° vs 25° tilt.

With the nrel-solar server connected, your agent calls compare_orientations and answers with production figures for both tilts. Every tool returns the same envelope — data + units + source + assumptions + warnings — so the agent always knows what a number means, where it came from, and which defaults were injected on its behalf.

Related MCP server: Solar MCP

Quickstart

1. Get a free NREL API key — https://developer.nlr.gov/signup/ (or use DEMO_KEY to try it out — 10 requests/hour).

2. Add the server to your MCP client — for Claude Desktop, merge this into claude_desktop_config.json (full example in examples/):

   {
     "mcpServers": {
       "nrel-solar": {
         "command": "uvx",
         "args": ["--from", "solar-mcp-nrel", "nrel-solar-mcp"],
         "env": { "NREL_API_KEY": "YOUR_KEY_HERE" }
       }
     }
   }

3. Restart your client and ask — try the Mesa tilt comparison above.

Each server is one console script: nrel-solar-mcp, solar-economics-mcp, solar-market-mcp, solar-forecast-mcp. Keys: NREL for production/irradiance, OpenEI + EIA for economics (all free); USPVDB and the forecast model need none. Verify keys and connectivity anytime:

$ uvx --from solar-mcp-core solar-mcp doctor

Development

$ git clone https://github.com/hoodsy/solar-mcp && cd solar-mcp
$ uv sync                # install the workspace
$ uv run pytest          # fixture replay only, no network

Layout: packages/core (shared HTTP client, cache, envelope) plus one package per server. Full spec: docs/SPEC.md. Smallest possible client: examples/example_client.py.

License

MIT. Per-source data licensing/attribution is exposed as MCP resources (source://<name>/license).