Connectors

A [Model Context Protocol](https://modelcontextprotocol.io) server that exposes the `international-criminal-tribunals-skills` suite to any MCP client (Claude Desktop, Claude Code, or another host). This is the first MCP server from the Impact Litigation Lab. It reads the tribunal skill folders in this repository live — it does not duplicate their content — and surfaces four capability areas that map onto the suite's verification-first methodology: | Area | Tool / resource | |---|---| | Expose the skills | `list_tribunals`, `get_skill_file`, `get_foundational_texts`, `skill://…` resources | | Verify citations | `verify_citation` | | Search the case law | `search_jurisprudence` | | Retrieve documents | `fetch_document` (with PDF text extraction) | The server never authorises citing case law from memory. `verify_citation` returns the *guidance* to verify a citation against a Tier 1 source; it does not assert correctness. `fetch_document` honours the suite's fallback ladder when an authoritative domain blocks a direct fetch (e.g. icc-cpi.int 403s).

PII Shield is an open-source MCP server for Claude Desktop (Cowork and Code) that anonymizes documents *before* they reach the model and restores real values *after* analysis — entirely on the user's machine. The pipeline is local → API → local. The plugin reads a file on disk, replaces detected entities with placeholders like `<PERSON_1>`, `<ORG_2>`, `<IBAN_1>`, and only the placeholder text crosses the wire. Detection combines GLiNER zero-shot NER with **17 jurisdiction-specific patterns** (UK NIN/NHS/passport, DE Tax ID, FR NIR, IT fiscal code, ES DNI/NIE, CY TIC, EU VAT, plus US SSN/EIN, IBAN, crypto wallet, money, dates). A human-in-the-loop review UI renders inside Claude Desktop for false-positive correction. Ships with a companion skill `pii-contract-analyze` exposing six modes — MEMO, REDLINE, SUMMARY, COMPARISON, BULK (up to 5 files, shared placeholders), and ANONYMIZE-only. **Jurisdictions covered (detection patterns):** EU-wide, UK, Germany, France, Italy, Spain, Cyprus, US. Use PII Shield whenever client documents must be analyzed by an LLM but cannot be transmitted in cleartext: NDA / MSA / DPA review, GDPR risk analysis, due diligence on counterparty contracts, employment document handling, DSAR responses, and partial-analysis handoff between colleagues under privilege.

MCP server for German federal laws, EU regulations/directives, and German federal court decisions — hybrid semantic + BM25 search across the full corpus, with verified citation lookup, table-of-contents navigation, and norm-to-case-law tracing. Lawbster makes German and EU primary law machine-readable for AI assistants via the Model Context Protocol. The server indexes the full body of German federal law (Gesetze-im-Internet — thousands of statutes and regulations), EU regulations and directives (EUR-Lex), and the full case-law of all German federal supreme courts and the EU courts. Lawbster isn’t just another AI tool built in isolation. The underlying logic, the compliance checks, and the regulatory frameworks it draws on were shaped with input from experienced lawyers at Planit // LEGAL a leading IT and data protection law boutique — practitioners who work with these regulations day to day, not just in theory. Search uses a three-stage hybrid pipeline (dense vector + BM25 sparse → RRF fusion → cross-encoder reranking), so colloquial questions resolve against statutory language. Citations like § 823 BGB, Art. 6 DSGVO, or C-311/18 resolve to verified full text — never hallucinated. The norm → Rechtsprechung chain surfaces the controlling case-law line behind a statute, which is often more decisive than the bare statute text. The current launch corpus covers German federal law and EU law. Additional jurisdictions and source types are on the roadmap — UK law (UK Public General Acts, UKSC/EWCA case-law) is the next planned expansion, followed by further European jurisdictions (e.g. AT, CH, FR, NL). The MCP tool surface stays stable; new jurisdictions are added via the existing jurisdiction filter and source_type enum, so existing client integrations keep working.

Moonlit is a Dutch legaltech company with a mission to make global legal information accessible and structured. A team of data engineers and legal engineers aggregates legal data from official government sources worldwide and transforms it into structured, continuously updated legal intelligence. Trusted by ~6,000 users including judges, lawyers, and researchers. ISO 27001 certified, EU-hosted. Moonlit provides global legal data infrastructure for AI workflows. The platform covers 190+ jurisdictions with 150+ official legal sources, updated daily. It supports keyword, semantic, and hybrid search with reranking across 11M+ legal documents — case law, legislation, and regulatory texts. Use Moonlit for cross-border legal research spanning multiple EU and national jurisdictions. It is well-suited for regulatory monitoring, multi-jurisdictional due diligence, and building multilingual legal retrieval pipelines. Jurisdictions: EU (28+ jurisdictions), UK, France, Germany, Benelux, Switzerland, UN

Local stdio MCP server that connects AI assistants to the Légifrance API for searching French codes, legislation, decrees, Journal Officiel publications, and retrieving legal articles.

Local stdio MCP server that connects AI assistants to the Judilibre API for searching, retrieving, and analyzing French court decisions from the Cour de cassation and courts of appeal.

GoodLegal is an AI-powered legal research API built by Zacharie Laik, focused on French and EU law. Where Legal Data Hunter indexes global sources, GoodLegal provides a developer-ready API with sub-second response times for French court decisions, legal codes, and EU texts. OAuth and API key authentication.

Legal Data Hunter is a community-driven initiative with the goal of indexing every source of law on the planet into a single searchable layer. Built on the philosophy that the law belongs to everyone, the platform crawls official government sources and adds new ones approximately every 30 minutes. Contributors can suggest sources, vote on priorities, and apply as Jurisdiction Leads. Legal Data Hunter gives your AI assistant access to 18.6M+ legal documents across 108 countries and 533 official government sources. It covers 13.4M+ court decisions, 4.6M+ laws and regulations, and 637K+ doctrine texts. The MCP server supports hybrid semantic + keyword retrieval with country, source, and court-level filtering — the broadest multi-jurisdictional legal coverage available via MCP. Use Legal Data Hunter for large-scale multi-jurisdictional legal search, legal citation resolution, or comparative law research. It excels at due diligence workflows and building legal knowledge pipelines that need the broadest possible country coverage. Jurisdictions: 108 countries (13.4M+ decisions, 4.6M+ laws, 637K+ doctrine texts)

Adeu acts as a "Virtual DOM" for Microsoft Word documents. It provides a two-way abstraction layer that lets AI agents freely edit DOCX files (Track Changes, Comments, Accept/Reject) via a lightweight Markdown proxy. This allows LLMs to negotiate and redline contracts without destroying the underlying OOXML formatting, images, or tables. On Windows, it natively connects to the active MS Word application to perform edits live on the canvas.

Deterministic legal tools for the EU AI Act. Covers Annex III high-risk system classification, Article 73 compliance deadlines, Article 99 fine calculations, and jurisdictional mapping across the 27 EU member states. 9 tools, public, no auth required.

data.gouv.fr is the French government official open data platform, operated by the DINUM (Direction interministerielle du numerique). This MCP server is maintained by the data.gouv.fr team itself — a first-party integration giving AI agents direct read-only access to France national open data catalog. MIT-licensed, actively maintained, no authentication required. Official MCP server for [data.gouv.fr](https://www.data.gouv.fr), the French government open data platform. Exposes **9 tools** across three categories: - **Dataset tools** (5): search datasets, get metadata, list resources, query tabular data in-place - **Dataservice tools** (3): discover APIs, retrieve OpenAPI specs - **Metrics tools** (1): track visits and downloads Covers 74K+ datasets, 387K+ resources, and 6K+ organizations. 1,124 GitHub stars, MIT license. Use DataGouv when you need to explore or analyze French public datasets — statistics, government reports, regulatory data, or open data catalogs. Particularly useful for legal and regulatory analysis that requires official public data sources. Jurisdictions: France (government open data)

OpenLegi is built by Le Lab at dassignies.law, a French law firm. It wraps the official PISTE Legifrance API — the French government legal data gateway — into an MCP server, giving AI assistants direct access to authoritative French and EU legal sources. Free tier available with shared rate limits; bring your own PISTE credentials for unlimited access. OpenLegi is an MCP server for official French legal sources, built on the Legifrance PISTE API. It exposes **30+ tools** across three service endpoints: - **Legifrance** (12 tools): search codes, jurisprudence, CNIL decisions, JORF, conventions collectives - **RNE/INPI** (9 tools): company search, BODACC, KBIS extracts, financial statements - **EUR-Lex** (9 tools): EU law search, article retrieval, metadata Covers 73 legal codes, 1.2M+ court decisions, and 14 active databases. Use OpenLegi when you need access to authoritative French legal sources — official codes, consolidated legislation, published case law, CNIL decisions, or collective agreements. Use the RNE service for company due diligence and the EUR-Lex service for EU law research. Jurisdictions: France (Legifrance, CETAT, CONSTIT, CNIL, KALI, JORF, BOFiP, BODACC), European Union (EUR-Lex, subscription)