Give AI agents structured access to
U.S. climate data.
Explore the MCP tools AI agents receive, the JSON returned by the server, and a complete tool-to-answer example.
- 1 You ask a question
- 2 The AI agent picks a tool
- 3 We return JSON
- 4 The AI agent writes the answer
The data retrieval happens on our side. The MCP client decides when to call a tool and how to use the result.
How AI Agents Use Climate Data
Together, these tools let AI agents find the right station, check available datasets, retrieve climate normals, and understand the local climate.
This is the list a connected client receives, with the descriptions the model reads when it decides which tool to call.
search_locations
read-only
Find U.S. climate stations by city name. Start here when you only know a place name: every other tool needs a station ID.
station_capabilities
read-only
Check which climate datasets exist for a station (monthly normals, daily normals, extended normals, snow, sunshine, wind) before requesting them. Use this when a station returns less data than expected, to tell "no data recorded here" apart from "wrong station".
get_monthly_climate
read-only
Get a ready-to-use climate profile for a U.S. station, combining long-term monthly climate values with a written description of the local climate. This is the best fit for questions about what a place is typically like. For the underlying official NOAA 1991-2020 reference values, use get_monthly_normals instead.
get_monthly_normals
read-only
Get official NOAA 1991-2020 monthly climate normals for a U.S. station, including temperature, precipitation, and snow. Use get_monthly_climate instead when you want a ready-to-use climate profile with contextual description.
get_daily_normals
read-only
Get official NOAA 1991-2020 daily climate normals for a U.S. station: what a given calendar day is typically like. Requires the Developer plan.
get_koppen_classification
read-only
Get the Köppen-Geiger climate classification for a U.S. station: the code (e.g. Csa), its group, name and what it means. Answers "what kind of climate is this" in one call, without deriving it from monthly figures.
Every call, in full
Pick a station and a tool. The left panel is what the model sends, the right is what the server returned when this page was recorded.
The climate description ships with the dataset. It is not generated by the assistant at request time.
Watch AI agents use it
Pick a question and press Run. Each tool call opens to the JSON behind it.
The tool calls and returned data are real. Only the final assistant wording is pre-written for this demo.
Build with U.S. Climate Data
Use the REST API today for direct integrations. MCP access is included with the Developer plan and is currently being rolled out for AI agents and agent workflows.