Judicex
github.com
What it is
Judicex is an open-source legal AI workspace for matter analysis, evidence-grounded drafting, and review. Its workflow packs compare a legal thesis with matter materials, identify elements that are present, partial, or missing, and suggest next actions. Users can upload PDF, DOCX, text, Markdown, CSV, JSON, and image files, with OCR via Ollama. Drafting features include templates, a split-view page, tabular review, and DOCX or PDF export. The answer engine ties citations to retrieved legal evidence and is designed to abstain if evidence is insufficient. It supports Ollama, OpenAI, Anthropic, OpenAI-compatible providers, and MCP-aware clients, and can run without an LLM. Matter data is stored in an operator-controlled SQLite database and attachment directory, remaining on the host unless a hosted LLM is configured. The project lists a local password gate, backup and restore, and optional matter-file encryption. Its v0.2.0-alpha release is described as a prototype for local development and private evaluation, not production SaaS; it offers no security guarantee for shared, multi-user, or internet-exposed deployments.
Spec sheet
- Product
- Judicex describes itself as an open-source legal AI workspace for evidence-grounded drafting, matter analysis and verifiable answers.Read 9 Oct 2026 · github.com
- Intended users
- The project describes its workspace as intended for lawyers and legal teams.Read 9 Oct 2026 · github.com
- Platforms
- The repository provides script-based local installers for Windows and macOS, and describes running the web app on a workstation or private server.Read 4 Oct 2026 · github.com
- Integrations
- The software supports Ollama, OpenAI, Anthropic and OpenAI-compatible LLM providers, as well as an MCP stdio server for MCP-aware clients.Read 9 Oct 2026 · github.com
- Privacy
- Matter data is stored in a SQLite database and attachment directory controlled by the operator, and data stays on the host unless the operator configures a hosted LLM.Read 4 Oct 2026 · github.com
- Support
- The security policy directs vulnerability reports through GitHub Security Advisories or the maintainer contact channel and says to expect acknowledgement within a reasonable time.Read 9 Oct 2026 · github.com
- Compliance limits
- The security policy says deployments requiring audited infrastructure for requirements such as HIPAA or FINRA are not supported without significant additional engineering.Read 4 Oct 2026 · github.com
- Documents
- Users can upload PDF, DOCX, text, Markdown, CSV, JSON and image files, with OCR via Ollama.Read 9 Oct 2026 · github.com
- Drafting
- Judicex provides a split-view drafting page and can export drafts to DOCX or PDF.Read 9 Oct 2026 · github.com
- Drafting and review
- It includes document templates, a split-view drafting page, tabular review, and DOCX/PDF export for drafts.Read 4 Oct 2026 · github.com
- Evidence and answers
- Its answer engine binds citations to retrieved documents and can abstain when evidence is insufficient.Read 9 Oct 2026 · github.com
- Evidence and citations
- Its answer engine binds citations to retrieved legal evidence and is designed to abstain when evidence is insufficient.Read 4 Oct 2026 · github.com
- Installers
- Script-based local installers are documented for Windows and macOS, and launch Judicex at a localhost address.Read 9 Oct 2026 · github.com
- Legal disclaimer
- Judicex states that it does not provide legal advice or replace a qualified legal professional, and that outputs may be incomplete, incorrect, outdated or unsuitable for a jurisdiction.Read 9 Oct 2026 · github.com
- Local storage and privacy
- Matter data is stored in a SQLite database and attachment directory selected by the operator, and data stays on the host unless a hosted LLM is configured.Read 9 Oct 2026 · github.com
- Matter analysis
- Matter analysis maps a legal thesis to proof elements and reports present, partial or missing support with next actions.Read 9 Oct 2026 · github.com
- Product maturity
- The repository labels the current release v0.2.0-alpha and describes it as a usable prototype rather than production SaaS.Read 4 Oct 2026 · github.com
- Security controls
- The project says API keys should be kept in environment variables and are not sent to the browser.Read 9 Oct 2026 · github.com
- Security features
- The listed privacy and operations features include a local password gate, backup and restore, optional matter-file encryption, and API keys kept out of the browser.Read 4 Oct 2026 · github.com
- Security limitations
- The current v0.2.0-alpha is intended for local development and private evaluation, and the project gives no security guarantee for shared, multi-user, or internet-exposed deployments.Read 4 Oct 2026 · github.com
- Security limits
- The security policy says Judicex v0.2.0-alpha has no security guarantee for shared, multi-user or internet-exposed deployments and is not designed for audited regulatory workloads without significant additional engineering.Read 9 Oct 2026 · github.com
- Support and reporting
- The security policy asks users to report vulnerabilities privately through GitHub Security Advisories or a maintainer contact channel listed on the GitHub profile.Read 4 Oct 2026 · github.com
Strong points and limits
Strong points
- Citations are tied to retrieved legal evidence.
- Can abstain when evidence is insufficient.
- Supports several LLM providers and no-LLM use.
- Matter data stays on the host unless hosted LLMs are configured.
- Exports drafts to DOCX or PDF.
Limits
- Current release is v0.2.0-alpha prototype software.
- No security guarantee for shared or internet-exposed deployments.
- Audited regulatory workloads need significant additional engineering.
- Outputs may be incomplete, incorrect, outdated, or jurisdictionally unsuitable.
Who it suits
Judicex is intended for lawyers and legal teams, including a single lawyer or small team running it locally or on a private firm server. Its stated prototype status and security limits make deployment context an important consideration.
Verdict
Judicex offers matter workflows, evidence-linked answers, and local data control, with support for multiple providers or no LLM. The project explicitly frames the current release as a prototype and warns against shared, internet-exposed, and audited-regulatory deployments without substantial additional engineering.
Where Specifiction places it
- AI Legal Research Software#19 of 327.6