JudicialMind
Resources

Open legal AI, built in the open.

The research behind JudicialMind does not stay locked up. We publish open-source skills and agents for legal AI, and large-scale legal datasets, so the wider community can build, evaluate, and push the field forward.

GitHubOpen-source plugin

Legal Skills

A comprehensive legal-practice skills library for AI coding assistants, with expert-level guidance across 30 practice areas, functional domains, and jurisdictions. Works with Claude Code, OpenAI Codex, and any tool that supports skills or plugins.

30
Skills
3
Skill groups
40+
Sourced legal databases
  • 15 practice verticals: M&A, litigation, IP, real estate, employment, tax, immigration, criminal, family, bankruptcy, healthcare, environmental.
  • 10 functional domains: legal research, contract lifecycle, e-discovery, due diligence, compliance, case management, intake, analytics, billing, drafting.
  • 5 jurisdiction and specialty skills: US federal courts, India legal, UK and Commonwealth, international arbitration, regulatory compliance.
  • Three-level progressive disclosure keeps token usage low while loading deep reference material only when a skill activates.
Claude CodeOpenAI CodexMITSkills
View on GitHub
GitHubOpen-source framework

The Legal Agency

An open-source, multi-agent framework for AI-native legal practice. 30 specialized agent definitions across three divisions, each a self-contained file with structured identity, workflows, deliverables, and zero-hallucination guardrails.

30
Agents
3
Divisions
MIT
License
  • 12 practice-vertical agents, each a senior specialist with deep statute, regulation, and case-law knowledge.
  • 10 functional-domain agents for research, contracts, e-discovery, analytics, billing, and document drafting.
  • 8 jurisdiction and specialty agents spanning US, India, UK, and international arbitration.
  • Platform-agnostic markdown that drops into orchestration frameworks, Copilot agent mode, RAG pipelines, or prompt libraries.
Multi-agentNode.js SDKMITAgents
View on GitHub
Hugging FacePublic dataset

Legal Training Dataset

A large-scale, multilingual query and passage corpus for training and evaluating legal information-retrieval and question-answering systems, with rich per-row metadata and a clean train, validation, and test split.

3.69M
Query and passage pairs
35
Languages
2.6 GB
264 parquet files
  • Spans Asia, Europe, North and South America, and Oceania, with jurisdiction labels per row.
  • Per-row metadata: query type, legal domain, difficulty, token count, language, country, jurisdiction.
  • File-level A, B, C bucket split keeps proportional coverage across languages and domains.
  • Built to fine-tune dense retrievers and rerankers, train multilingual legal QA models, and benchmark RAG pipelines.
RetrievalQAMultilingualCC-BY-NC-ND
View on Hugging Face
Hugging FacePublic dataset

India Acts: Central and State

A comprehensive corpus of Indian legislation in PDF form, covering both Central Parliament Acts and State and Union Territory Acts in English and Hindi, consolidated from publicly available government sources.

12,102
Act PDFs
21.7 GB
Total corpus
36
States and UTs
  • 1,593 Central Acts in English and Hindi, spanning 1836 to 2025.
  • Over 10,000 State and Union Territory Acts across all 28 states and 8 UTs.
  • Bilingual coverage with strong Hindi representation across the Hindi-belt states.
  • Built for legal retrieval, statutory question-answering, summarization, OCR and parsing benchmarks, and multilingual legal NLP.
IndiaStatutesBilingualPDF corpus
View on Hugging Face