Searches IPCC AR6 WG1 material and returns page-cited answers, refusing questions it cannot support.
Copy the AI prompt to install this server into Claude Code, Cursor, or another agent — or use 1-click editor setup below.
One-click editor setup isn’t available for this listing yet — we don’t have a confirmed install command, and we’d rather show nothing than point your editor at the wrong package or host. Follow the project’s own setup instructions, linked above.
Inspect callable tools, capabilities, and parameters exposed to AI agents by IPCC AR6 Climate Retrieval.
IPCC AR6 Climate Retrieval MCP server retrieves evidence from the IPCC Sixth Assessment Report Working Group I corpus and uses it to answer climate-related questions. Answers include the page numbers used as supporting citations. If the system cannot locate evidence for a claim in the retrieved material, it refuses instead of filling the gap with an unsupported answer.
The repository describes a climate-risk analyst workflow rather than a general-purpose document search interface. A question can include a place name or direct latitude and longitude, along with a hazard such as heat, extreme rain, or wind. The resulting report can place a forecast beside more than 60 years of ERA5 annual-maximum data and an IPCC AR6 projection for the applicable reference region.
Input first passes through parsing, geocoding, and mapping to an IPCC AR6 reference region. The agent then follows a five-node LangGraph sequence: planning, forecast calls, research, Chapter 12 projection extraction, and report synthesis. The output is a typed RiskReport, or a typed refusal when the question is outside the supported scope.
IPCC AR6 retrieval uses BM25 and dense retrieval combined through reciprocal-rank fusion. Retrieved citations are checked against the pages actually returned for the question. Chapter 12 climatic impact-driver projections are extracted from retrieved chunks, including the quoted sentence, direction of change, calibrated confidence wording, and the stated warming level or period when available.
The location analysis fits stationary and drifting-location Generalized Extreme Value models to ERA5 annual maxima. A likelihood-ratio test determines whether a warming trend is supported. Return levels include 90% bootstrap confidence intervals, and the risk band is based on the forecast peak's estimated return period at that location. The documented bands are below two years for low, two to ten years for moderate, ten to 50 years for high, and at least 50 years for severe.
The documented hazard scope centers on heat, extreme rain, and wind. Questions outside the system's checked scope may receive a refusal. Regional projections are not shown when the corpus contains no assessed change for the requested region and hazard.
Evaluation results in the README are measurements for the repository's stated configuration, datasets, and test questions; they are not guarantees for every query. The held-out evaluation reports zero false answers, 96% citation validity, and 88% numeric provenance, while the broader development results differ. The README does not provide an MCP package name, a direct installation command, required credentials, or client-specific configuration. It also does not document a standalone tool schema beyond the described climate-risk workflow.
Factual signals from GitHub, npm, and our automated checks — not a rating.
No reviews yet — be the first to share how this listing worked for you.
Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.
[](https://allmcps.com/mcp/ipcc-ar6-climate-retrieval)<a href="https://allmcps.com/mcp/ipcc-ar6-climate-retrieval"><img src="https://allmcps.com/api/badge/ipcc-ar6-climate-retrieval?style=directory" alt="IPCC AR6 Climate Retrieval on AllMCPs" /></a>