Open Data, API and MCP

Open Legal Data collects German court decisions and laws from official sources, extracts metadata and the citations between them, and publishes everything as open data. Pick the access that fits your use case:

MCP server for AI assistants

With the Model Context Protocol, AI assistants can work directly with Open Legal Data: search court decisions and laws, read full texts, validate citations and follow the citation graph between cases and laws. Answers can link to the sources, so you can check them.

Server URL:

https://de.openlegaldata.io/mcp

• claude.ai and Claude Desktop: add a custom connector under Settings → Connectors and enter the server URL.
• Claude Code: claude mcp add --transport http oldp https://de.openlegaldata.io/mcp
• ChatGPT: add a custom MCP server in the browser on chatgpt.com (depending on your plan); the step-by-step guide is on our API page and in the ChatGPT help.
• Other clients (Cursor, your own agents, …): add the URL as a “Streamable HTTP” server. Clients without OAuth can send an API token in the Authorization: Token … header.

The server works without an account. Signing in via OAuth (or using an API token) gives you a personal, higher request budget, shared with the REST API.

Example prompts:

• “Which decisions of the Federal Court of Justice since 2020 deal with § 823 BGB and personality rights?”
• “Summarise the decision [file number] and list the laws it cites.”
• “Check whether these citations exist: [file numbers, ECLI or sections]”

All tools and parameters are described in the MCP documentation.

REST API

The API at https://de.openlegaldata.io/api/ returns JSON (or XML with ?format=xml). Try requests in the Swagger UI or start with:

# Decisions mentioning "Mietminderung" (rent reduction)
curl "https://de.openlegaldata.io/api/cases/search/?text=Mietminderung"

# Decisions citing § 823 BGB that mention "Persönlichkeitsrecht"
curl -G "https://de.openlegaldata.io/api/cases/search/" \
  --data-urlencode "text=Persönlichkeitsrecht" \
  -d cited_law_book=bgb -d cited_law_section=823

Requests work without an account. With a free account and an API token the limits are considerably higher. Endpoints, limits and token setup are documented on the API page (in German); developer documentation is on openlegaldata.github.io/oldp. Please do not page through the API to download everything: use the datasets instead.

Datasets

We publish regular snapshots of the database on Hugging Face. Each dataset is versioned per dump and has random 1k and 10k samples for quick experiments.

from datasets import load_dataset

cases = load_dataset(
    "openlegaldata/court-decisions-germany",
    name="dump-20260520-1k",  # see the dataset card for the latest dump
    split="train",
)

The dataset cards list the available dumps; the release posts on our blog have statistics per dump. To process the raw data yourself, use the oldp-toolkit.

License

In Germany, laws, official decrees and court decisions are not protected by copyright (§ 5 UrhG). The collection and its metadata are released under the Open Database License (ODbL) v1.0. If you use the data in research, please cite our paper.

How it works

Open Legal Data harvests legal documents from official sources, extracts information from the raw documents and enriches them with data from additional sources. The content is available through the website, the API, the MCP server and the datasets.

All code is open source on GitHub under the MIT license. Questions or ideas? Join us on Discord.