Live · Climate data, ready to use
Your agent has a question.Give it better data.
Climate data for AI agents, through an API or MCP. Specify the place, period and format. Verdant retrieves, normalizes and delivers the data with its evidence intact.
01The problem
Climate data arrives in every shape: grids, stations, spreadsheets, archives. Units drift. Resolutions clash. Calendars disagree. Your agent shouldn't have to untangle any of it. Verdant retrieves, normalizes and delivers exactly what was asked for, with its evidence intact.
Hi! Ask me about Verdant's climate and farm data.
02Playground
Ask in plain language. Get the evidence.
Verdant's analyst finds the published datasets that fit, works through them in a sandbox and writes up its method with charts, equations and downloadable tables. Type a question, or talk it through with Sprout.
- Did the CSIRO vineyard near Mildura irrigate more during the hottest spells of summer 2003–04?
- What were the hottest and coolest days near Mildura in January 2004?
- How reliable are the NWS day-two frost forecasts? Show their calibration.
- Did higher nitrogen rates pay off in the Ohio corn trials?
- 01Answer
- 02Data and provenance
- 03Methodology
- 04Charts
- 05Limitations
03How it works
From question to clean data.
Agents describe what they need. Verdant handles retrieval, reconciliation and checks, then hands back the exact representation requested, with nothing silently changed.
- 01
Request
Name the variable, place, period, units and format. Meaning stays separate from serialization, so “temperature as CSV” is never ambiguous.
Live - 02
Retrieve
Published data returns instantly. Anything not yet published is fetched from the source on request by an agent with narrow, audited tools, paid per acquisition through MPP.
Live - 03
Normalize
Units, calendars, spatial support and missing values are reconciled and documented, never quietly relabeled or filled.
Live - 04
Validate
Deterministic checks reconcile every cell with the source bytes and confirm dimensions, units and coverage before anything is published.
Live - 05
Deliver
Receive JSON values with provenance directly, or request CSV in the response. Pin the dataset version for reproducible queries.
Live
04Provenance
Evidence intact.
Every delivery carries the meaning of its numbers: where they came from, what they measure and what was done to them. Your agent can cite it and check it.
- Never
Missing values → zero - Never
Coarse pixel → fine measurement - Never
Undocumented interpolation
Supported requests
- Sources
- SILO · NOAA nClimGrid · NOAA CPC
- Variables
- Max & min temperature · precipitation
- Units
- °C · mm
- Resolution
- Native grid · daily
- Coordinates
- WGS84 · EPSG:4326
- Data class
- Interpolated observation
- Missing data
- Preserved
- Formats
- CSV · JSON
05Try the contract
One clear contract. Every source accounted for.
Query daily maximum temperature around Mildura, Australia. Published dates return instantly with their source and version; Australia, the US and global land can be requested for any covered date.
- Source
- SILO
- Units
- Degrees Celsius
- Resolution
- Native grid · daily
- Missing data
- Preserved
06Build with Verdant
Read the docs. Ship the query.
Reading published data is free and needs no API key. New data is acquired on request and paid per acquisition with MPP. Start over HTTP, or hand your agent the MCP server.
- 5 min
Quickstart
Discover coverage and get real values directly in an HTTP response.
- MCP
https://api.verdant-ai.com/mcpConnect an agent
Add the Verdant MCP server to any client that supports remote MCP.
- Concepts
The data contract
Units, spatial support, missingness and the guarantees behind every response.
- OpenAPI 3.1
API reference
Every endpoint, generated from the same runtime schemas the API validates with.