OpenClimateLaw

A citation-safe Model Context Protocol layer over the world's climate litigation and legislation. Built so AI agents (Claude, ChatGPT, Gemini, Copilot, Cursor and others) can quote and cite climate cases and laws without fabricating either.

Connect in a minute

No sign-in and no API key. Pick your assistant and follow the steps — then ask it a climate-law question.

Toolkit for climate litigators

OpenClimateLaw is part of a broader regenerative.law commitment to free, open-source legal-research infrastructure for the climate-litigation field. Our intended audience is activist lawyers, university clinics, public-interest organisations, and researchers — practitioners who do this work without budgets for commercial AI research tools.

regenerative.law — an initiative of regenerative Foundation — supports climate litigation in two complementary ways: by funding cases directly, and by operating shared infrastructure that lowers the cost of doing this work. The tool below is run as a public service — free, unauthenticated, open source, with attribution preserved to the upstream sources we redistribute.

More tools are in development. If you operate a public-interest organisation in the climate-litigation space and would benefit from infrastructure we don't yet provide, please write to us.

What it does

Seventeen research tools in six groups — plus search and fetch, the two read-only tools ChatGPT's connector calls (19 in total). Each is callable from any MCP client; combine them for full-fidelity legal research.

Try it

Once connected, ask any compatible client:

• "Search the climate-litigation corpus for 'Amazon deforestation' filed in Brazil; show the top five with their core_object."
• "Find Urgenda Foundation v. State of the Netherlands, cite it in en/sabin format, and quote one sentence from its summary verbatim — but verify the quote first using check_claim_support."
• "Find youth-plaintiff cases in the United States and summarize their claim types."
• "Search the climate-legislation corpus for the UK Climate Change Act 2008 (search_statutes), then quote its emissions-target language verbatim from get_statute."
• "Get Friends of the Earth and Others v. Secretary of State (GB): show its parties and the CCLW laws it references (linked_statutes), then pull the full text of the Climate Change Act 2008."
• "Show the newest climate decisions in the United States from 2025 onward — sort by decision date and list each case's decision_date."
• "How many climate cases are filed against governments per jurisdiction? Top 10."
• "Validate this draft: 'In ECLI:NL:HR:2019:2007 the court reasoned, citing Massachusetts v. EPA, 549 U.S. 497 (2007)…' against retrieved cases [urgenda-foundation-v-state-of-the-netherlands]."

Add to Claude

Claude Code (CLI)

claude mcp add openclimatelaw https://mcp.openclimatelaw.org/mcp --transport http

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows). Uses mcp-remote as a stdio→HTTP bridge.

{
  "mcpServers": {
    "openclimatelaw": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.openclimatelaw.org/mcp"]
    }
  }
}

claude.ai (Free / Pro / Max / Team / Enterprise)

Customize → Connectors → click + → enter a name and the URL https://mcp.openclimatelaw.org/mcp → Add. (Connectors recently moved from Settings to Customize; available on Free — one connector — through Enterprise.)

Anthropic API (Messages, MCP Connector)beta

import anthropic

client = anthropic.Anthropic()
resp = client.beta.messages.create(
    model="claude-opus-4-8",
    max_tokens=1024,
    betas=["mcp-client-2025-11-20"],
    mcp_servers=[{
        "type": "url",
        "url": "https://mcp.openclimatelaw.org/mcp",
        "name": "openclimatelaw",
    }],
    tools=[{"type": "mcp_toolset", "mcp_server_name": "openclimatelaw"}],
    messages=[{"role": "user", "content": "How many climate cases involve Brazil?"}],
)

Add to ChatGPT & OpenAI

ChatGPT (developer mode — Plus / Pro / Team / Enterprise)

Settings → Apps & Connectors → Advanced settings → enable Developer mode, then Create a connector with the URL https://mcp.openclimatelaw.org/mcp and Authentication None. Enable it in a chat via + → More. If a flow (e.g. Deep Research) asks for an SSE URL instead, use https://mcp.openclimatelaw.org/sse — both serve the same tools, including the search and fetch tools ChatGPT calls. (Custom connectors are unavailable on the Free plan.)

OpenAI Agents SDK (Python)

from agents import Agent, Runner
from agents.mcp import MCPServerStreamableHttp

async with MCPServerStreamableHttp(
    name="openclimatelaw",
    params={"url": "https://mcp.openclimatelaw.org/mcp"},
) as server:
    agent = Agent(name="research", mcp_servers=[server])
    result = await Runner.run(agent, "Find Urgenda v. Netherlands and cite it.")
    print(result.final_output)

OpenAI Responses API (mcp tool)

from openai import OpenAI
client = OpenAI()
resp = client.responses.create(
    model="gpt-5",
    tools=[{
        "type": "mcp",
        "server_label": "openclimatelaw",
        "server_url": "https://mcp.openclimatelaw.org/sse",
        "require_approval": "never",
    }],
    input="Look up the Urgenda case and quote one sentence verbatim, validated.",
)

Add to GitHub Copilot

VS Code (workspace)

Add to .vscode/mcp.json:

{
  "servers": {
    "openclimatelaw": {
      "type": "http",
      "url": "https://mcp.openclimatelaw.org/mcp"
    }
  }
}

Copilot CLI

copilot mcp add --transport http openclimatelaw https://mcp.openclimatelaw.org/mcp

(Exact subcommand depends on your Copilot CLI version; consult copilot mcp --help.)

Add to Google Gemini

The consumer Gemini app (gemini.google.com) has no custom-connector UI yet — connect via the CLI, the SDK, or Gemini Enterprise below.

Gemini CLI

gemini mcp add --transport http openclimatelaw https://mcp.openclimatelaw.org/mcp

Or edit ~/.gemini/settings.json directly with an mcpServers block (same shape as Claude Desktop).

Google AI Studio / Vertex AI via tools API

Gemini's tool-use API can call MCP servers via the SDK's MCP client integration:

import asyncio
from google import genai
from mcp import ClientSession
from mcp.client.streamable_http import streamablehttp_client

async def run():
    async with streamablehttp_client("https://mcp.openclimatelaw.org/mcp") as (read, write, _):
        async with ClientSession(read, write) as session:
            await session.initialize()
            client = genai.Client()
            resp = await client.aio.models.generate_content(
                model="gemini-2.5-pro",
                contents="How many climate cases involve Germany? Use the MCP tools.",
                config={"tools": [session]},
            )
            print(resp.text)

asyncio.run(run())

Other MCP clients

Any MCP client supporting Streamable HTTP works. Point it at https://mcp.openclimatelaw.org/mcp.

• Cursor: Settings → MCP → Add new MCP server → URL transport.
• Continue.dev: add to ~/.continue/config.yaml under mcpServers with type: streamable-http.
• Cline (VS Code): MCP Servers panel → Edit settings → add a URL entry.
• Goose: goose configure → Add MCP Server → choose SSE or HTTP.
• Zed: add to settings.json under "context_servers".
• Custom code: any HTTP client speaking the MCP Streamable HTTP spec.

Data sources, methodology & licence

Centred on the Sabin Center for Climate Change Law's Climate Litigation Database (4,812 cases, full metadata + summaries + court-document text), scraped directly from www.climatecasechart.com under the CC-BY 4.0 licence Sabin publishes the data under. Court-document PDFs are downloaded only from climatecasechart.com/wp-content/uploads/… on Sabin's own domain. No third-party API in the loop.

Metadata-only references to the Climate Rights Database (215 cases) are included with redirect-to-source links. Full case summaries are not copied — consult CRD directly for substantive content.

The legislation layer is the Climate Policy Radar Climate Change Laws of the World corpus, published as open data (CC-BY) on Hugging Face — 5,347 laws & policies across ~200 jurisdictions, with full multilingual text. Cases are cross-linked to the laws they reference by exact title + jurisdiction (145 high-confidence links so far; broader linking is on the roadmap). Parties are parsed from case captions (92% of cases); outcomes are LLM-classified under a confidence gate (below).

Per CC-BY 4.0, every citation_string returned by this MCP carries upstream attribution. The MCP server itself is open source under the MIT licence at github.com/jonashertner/openclimatelaw.

Embeddings: sentence-transformers/all-MiniLM-L6-v2 (384-dim) stored as pgvector with HNSW indexing. Citation graph: Aho-Corasick canonical-title matching plus formal-cite regex extraction. Outcomes: classified by claude-sonnet-4-6, confidence-gated and verified against a verbatim quote — anchored to the latest disposition (so an appellate reversal isn't mislabeled by an overturned lower-court order) — 1,606 decided cases carry an outcome; the rest are left blank rather than guessed. Structured doctrine (disposition, holdings, legal test, bases, relief) is extracted by claude-sonnet-4-6 from each case's summary and decision text, with every quoted element verified verbatim against the source and flagged otherwise — a preview on 108 landmark cases so far (72% of holding-quotes verbatim-verified; the ICJ Advisory Opinion and Shell, 100%). Roadmap: bulk-export ingestion path with the data sources; broader case↔law linking via Climate Policy Radar's own concept IDs; scaling the doctrine layer across the decided corpus.

Source & status

Open source: github.com/jonashertner/openclimatelaw · MIT · v0.1 · MCP over Streamable HTTP (/mcp) + SSE (/sse) · Hetzner Cloud · Let's Encrypt automatic TLS.

Health check: https://mcp.openclimatelaw.org/health