Show HN: Judicex – Open-source legal AI that abstains instead of hallucinating

Category: devtools

Tags: legal-ai, evidence-grounded, rag

Score: 6.5/10 (Innovation: 6, Technical: 7, Documentation: 7, Utility: 6)

Judicex is an open-source legal AI workspace that emphasizes evidence-grounded answers to avoid hallucination, featuring a structured approach with versioned legal sources, deterministic workflows, and an 'answer contract' that abstains when evidence is insufficient. It provides a local-first stack with Flask UI, CLI, and MCP integration, supporting multiple LLM backends while keeping data under user control. The project is interesting for its principled design addressing a critical pain point in legal AI—trustworthiness and verifiability—and its dual evidence layer model.

Target audience: backend devs

Repository: https://github.com/JustVugg/judicex · Python · Apache-2.0 · 5 stars

View on Hacker News