Legal Data Hunter MCP server

World Wide Law

Open-source collection scripts for open legal data from 110+ countries.

Every country publishes its laws, court decisions, and regulations online -- but in different formats, behind different APIs, with different access rules. World Wide Law is building the open infrastructure to collect, normalize, and make all of it searchable.

All sources in this repository are open data -- publicly available legal information from official government portals, APIs, and bulk download endpoints. We always prefer API and bulk access over web extraction.

Live Dashboard & API

• Dashboard: legaldatahunter.com -- track coverage, explore sources, submit feedback
• Search API: Available at legaldatahunter.com -- search across 16M+ indexed legal documents

What's Here

This repository contains 960+ collection scripts across 110+ countries that download and normalize open legal data from government portals worldwide. Each script follows a standard interface so that any developer can run, test, or improve it. Some sources are marked as blocked (CAPTCHA, IP restrictions, etc.) -- their scripts are included so developers can review and potentially contribute fixes.

sources/
  FR/LegifranceCodes/     # French consolidated legal codes (API)
  DE/GesetzeImInternet/   # German federal laws (bulk XML)
  IT/NormattivaLegislation/ # Italian legislation (API)
  ES/BOE/                 # Spanish official gazette (API)
  ... (110+ countries)

Quick Start

# Clone the repo
git clone https://github.com/worldwidelaw/legal-sources.git
cd legal-sources

# Install dependencies
pip install -r requirements.txt

# Check project status
python runner.py status

# Test a specific source
python runner.py sample FR/LegifranceCodes

# See what needs work
python runner.py next

How It Works

Per-Source Structure

Every source lives in sources/{COUNTRY_CODE}/{SourceName}/ and contains:

Two Data Models

Legislation (mutable): Laws get amended. Same ID, new content. Strategy: upsert with version tracking.

Case law (immutable): Court decisions don't change after publication. Strategy: append-only with dedup.

Standard Output Schema

Every script normalizes documents to a common schema:

• _id -- Unique identifier
• _source -- Source identifier (e.g., FR/LegifranceCodes)
• type -- legislation or caselaw
• title -- Document title
• text -- Full text content
• date -- Publication or decision date
• url -- Link to the original source

Architecture

legal-sources/
  manifest.yaml          # Master inventory: all sources + status
  runner.py              # CLI: run, test, and manage collection scripts
  common/                # Shared libraries
    base_scraper.py        Base class all scripts inherit from
    http_client.py         HTTP client with retries + caching
    rate_limiter.py        Token bucket rate limiter
    storage.py             JSONL storage with deduplication
    validators.py          Schema validation
  templates/             # Templates for new sources
    scraper_template.py    Boilerplate for bootstrap.py
    config_template.yaml   Boilerplate for config.yaml
    retrieve_template.py   Boilerplate for retrieve.py
  sources/               # One directory per data source
    {CC}/{Source}/          (see per-source structure above)

Coverage

Shortened. The full README is on GitHub.