atmospore-mcp

MCP server for Atmospore. Lets Claude (and other AI assistants supporting MCP) answer questions about pollen forecasts at any point on Earth.

What it does

Adds four tools to Claude:

Tool	Use it for
get_pollen(lat, lon, forecast_days)	"What's the pollen forecast in Oslo this week?"
get_top_species(lat, lon, limit)	"What pollen is highest in Bergen right now?"
get_area_average(lat, lon, radius_km, forecast_days)	"Is tree or grass pollen worse in London this week?"
list_supported_species()	"What species do you cover?"

And one resource:

• atmospore://help — usage notes browsable from Claude Desktop

Coordinates work anywhere on the planet — the model has global coverage at ~28 km resolution.

Install

pip install atmospore-mcp

Setup in Claude Desktop

1. Get a free Atmospore API key at atmospore.com/account (100 calls/day, no credit card required).
2. Edit your Claude Desktop config:
   • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
   • Windows: %APPDATA%/Claude/claude_desktop_config.json
3. Add Atmospore as an MCP server:

{
  "mcpServers": {
    "atmospore": {
      "command": "atmospore-mcp",
      "env": {
        "ATMOSPORE_API_KEY": "ak_your_key_here"
      }
    }
  }
}

4. Restart Claude Desktop.
5. Ask Claude: "what's the pollen in Oslo today?" — it'll call the tool and report back.

Example prompts

• "What's the pollen forecast for Oslo this week?"
• "Is tree or grass pollen worse in Stockholm right now?"
• "I'm allergic to birch — when is birch pollen peaking in Bergen?"
• "Compare pollen levels in London and Copenhagen today."

Quotas

Each tool call hits the Atmospore API and counts against your key's daily quota. The free tier (100 calls/day) covers casual use — typical Claude conversations fire 1–5 tool calls. Heavy users (~5+ pollen conversations/day) will want the paid plan.

When you hit the quota the tool returns a structured quota_exceeded response and Claude will tell you in plain language: "You've hit your Atmospore daily limit. Upgrade at atmospore.com/plans."

Develop locally

git clone https://github.com/atmospore/atmospore-mcp
cd atmospore-mcp
pip install -e ".[test]"
pytest
ATMOSPORE_API_KEY=ak_... atmospore-mcp  # runs the server on stdio

Related

• atmospore — the underlying Python client this server wraps.
• atmospore.com — the hosted forecast and developer dashboard.

License

MIT.