LegalQuant AI (LQ-AI)

Open-source AI for legal teams. Bring your own keys, run it where you want, own your data.

License: Apache 2.0 PRD Status SLSA 3 Security Policy

LQ.AI is a self-hosted AI platform purpose-built for legal teams. It delivers conversational chat with persistent history and matter-scoped projects, character-verifiable citations against source documents (M2's four-stage Citation Engine), a privacy-preserving anonymization layer for cloud inference (M2), reusable workflow skills authored in the open agentskills.io / Anthropic Claude Skills format, and a curated library of starter skills for the everyday work lawyers actually do — running on a laptop, an internal server, or a cloud VM, against the customer's choice of model (Anthropic, OpenAI, Azure OpenAI, or local Ollama out of the box), with zero license fees.

Playbooks (codified legal positions for automated contract review) and multi-document tabular review shipped in M3, alongside a Microsoft Word add-in (installable/authenticated scaffold) and a Slack/Teams light-intake bridge (plumbing).

The Autonomous Layer (opt-in background agents under hard brakes) shipped in M4; the Contract Repository relationship graph remains roadmap. After M4, gateway-brokered legal research (case-law lookup via CourtListener) and operator-approved connectors (MCP) shipped under a governed tool-loop with a human confirmation gate — every external call routed through the Inference Gateway as the sole audited egress. See Project status below for the current shipped-vs-roadmap split.

The project's reason for existing is simple: legal teams should not have to choose between AI assistance and data sovereignty. Every other capable tool in this category is a closed-source SaaS that requires sending privileged information to a third-party vendor. LQ.AI runs in your environment, with your keys, against your choice of model — including fully air-gapped deployments using local inference.

📘 LQ.AI Core Capabilities Overview

LQ.AI is a specialized, governance-focused platform providing advanced AI capabilities for the legal domain. The system uses a modular, multi-layered architecture designed to maximize security, auditability, and operational efficiency across legal practices.

🧠 I. Conversational & Knowledge Layer (The Intelligence Core)

Conversational Core: Multi-turn chat with persistent context, history, and attachment support. Features include history searching and export functionality.

Knowledge Bases (KB): Centralized document repository. Supports hybrid retrieval (vector similarity + full-text search). Uses layout-aware parsing for complex PDFs, ensuring all context is integrated into the chat.

Skills Library: A repository of reusable, structured prompts (artifacts). Skills are categorized (built-in, user, shared) and are designed to be easily debuggable. Includes a Meta-Skill for conversational prompt creation.

Organization Profile: A singleton context layer that captures firm-wide standards, jurisdiction rules, and standard operating voice. Skills automatically adhere to this defined context.

Enhance Prompt: Optional pre-processing step that transforms simple queries into detailed, legal-grade prompts (defining role, scope, jurisdiction, etc.) to guarantee thoroughness.

🔎 II. Governance & Assurance Layer (Trust & Compliance)

Citation Engine: The system's core assurance feature. It verifies every model-generated citation against source documents using multi-stage checks (exact match, paraphrasing quality judge, ensemble verification). Citations are visually rendered with compliance states (Verified $\text{(Green)}$, Inferred $\text{(Yellow)}$, Unverified $\text{(Red)}$). Anonymization Layer: A mandatory pre-processing layer that detects and pseudonymizes sensitive PII (names, case IDs) before data leaves the secure environment for external model calls, ensuring privacy compliance.

Inference Tier Awareness: A persistent status indicator that dictates the data handling protocol (Tier 1 to 5). The platform enforces a minimum "floor" inference tier for every action, ensuring the required data confidentiality standard is met. Audit Log: An append-only, immutable ledger tracking every state-changing action. Logs include the effective inference tier, privilege status, and routing provider for full compliance traceability.

⚙️ III. Structured & Operational Layer (Process Management)

Projects (Matter-Scoped Containers): Allows users to segment all work (chats, files, skills, context) into specific matters or deals. Each container can be assigned a unique security and access profile, ensuring strict data isolation. Governed Workflows: Manages complex, multi-step processes (e.g., drafting a deal memo). Provides built-in compliance checks and mandatory checkpoints, ensuring the process is followed systematically and auditable. Client Workflow: Specific protocols for managing external communication and client data ingress/egress, maintaining clear boundaries and privilege markers throughout the lifecycle.

Key Benefits Summary

Data Sovereignty: Controlled by mandatory inference tiers and anonymization layers.

Risk Mitigation: Minimized via the auditable Citation Engine and structured Governed Workflows.

Efficiency: Achieved by centralized Knowledge Bases and reusable Skills Library.

Auditability: Complete, state-change logging via the Audit Log and citation verification.