🛡️ LexGuard — Multi-Agent AI Contract Intelligence

"Every contract is written by their lawyer. This is yours."

Built for the Google × Scaler PromptWars 2026 (Problem Statement 01). A state-of-the-art legal intelligence platform that breaks down complex contracts, exposes hidden risks, and equips you with custom-tailored pushback scripts.

---

📖 Executive Summary (Non-Technical)

The Problem

When you sign an NDA, employment agreement, or SaaS Terms of Service, you are signing a document drafted by their legal team to protect their interests. Hiring a lawyer costs upwards of $300/hour, leaving individuals and small businesses to sign high-risk agreements blindly.

The Solution

LexGuard levels the playing field. In under 30 seconds, it parses any contract, identifies hidden liabilities, and generates a comprehensive defence strategy.

📂 Upload Agreement ──► 🧠 3-Agent Gemini Chain ──► 📊 Interactive Risk Dashboard
(PDF/Text)               (Parser, Assessor, Defender)      (Scores, Audits & Pushbacks)

Key Capabilities

• ⚡ Instant PDF Extraction: Upload raw PDFs or paste plain text. LexGuard immediately digests the entire legal structure.
• 🚦 Tri-Color Risk Audit: Clauses are categorized dynamically:
  • 🔴 HIGH RISK: Extreme liabilities, indefinite survival, or one-sided waivers.
  • 🟡 MEDIUM RISK: Standard clauses with slightly unfavorable terms requiring caution.
  • 🟢 LOW RISK: Balanced, standard, or protective clauses.
• ⚖️ The Role-Aware Advocate: Highly customized pushback strategies based on who you are (e.g., an Employee will get different negotiation arguments than an Independent Contractor).
• 📁 History Vault: Securely preserves your past contract analyses in a central vault, allowing you to instantly reload, review, and export them at any time.

---

⚙️ Core Technical Architecture

LexGuard is built on a high-performance Next.js 16 (App Router) frontend containerized via Docker, continuously built by Google Cloud Build, and served worldwide on Google Cloud Run.

graph TD
    User([User Browser]) -->|1. Uploads PDF/Text| NextApp[Next.js App Router /app]
    NextApp -->|2. POST /api/analyze| API[API Route /api/analyze]
    
    subgraph Multi-Agent Gemini Chain
        API -->|3. Raw PDF/Text| Agent1[Agent 1: Structure Parser]
        Agent1 -->|Structured JSON| Agent2[Agent 2: Severity Assessor]
        Agent2 -->|Categorized Clauses| Agent3[Agent 3: Defender / Advocate]
        Agent3 -->|Final Clause & Pushbacks| API
    end
    
    API -->|4. Fire-and-Forget| Firestore[(Google Firestore)]
    API -->|5. Structured Analysis Result| NextApp
    
    classDef gemini fill:#4f46e5,stroke:#fff,stroke-width:2px,color:#fff;
    classDef database fill:#ea580c,stroke:#fff,stroke-width:2px,color:#fff;
    class Agent1,Agent2,Agent3 gemini;
    class Firestore database;

Loading

The 3-Stage AI Agent Pipeline

Instead of a single large query, LexGuard breaks the analysis down into three highly specialized serial AI agents powered by gemini-2.5-flash for surgical precision and maximum speed:

1. Agent 1 (The Parser): Extracts clauses from raw document text or parses Base64 PDF files natively using Gemini's multimodal processing. Standardizes output into a clean, verified structural layout.
2. Agent 2 (The Severity Assessor): Evaluates every extracted clause against a rigorous matrix. It ranks severity (HIGH, MEDIUM, LOW), highlights specific red flags, and details the core legal reasoning behind the rating.
3. Agent 3 (The Defender / Advocate): Triggered dynamically for HIGH RISK clauses. It analyzes the user's role (e.g. Employee, Contractor, Founder) and generates:
   • What it costs you: A plain-English explanation of the financial or legal liability.
   • Advocate Pushback Script: An exact, professional statement you can copy-paste to negotiate the clause.
   • Proposed Amendment: Redlined text ready for the modified agreement.

---

🛠️ The Technology Stack

Component	Technology	Google/GCP Services	Purpose
Core Framework	Next.js 16 (App Router) + React 19	-	High-performance, SEO-optimized, static & dynamic route rendering
Styling Engine	Tailwind CSS v4 (PostCSS)	-	Zero-runtime CSS utility engine with custom variables
Core Intelligence	Gemini 2.5 Flash	Google AI Studio	Multi-agent reasoning, PDF extraction, and legal synthesis
Database	Firebase Admin SDK	Google Cloud Firestore	Secure serverless storage for analysis logs and history vault
Deployment	Docker Standalone Container	Google Cloud Run	Zero-scaling serverless edge hosting
CI/CD	Git Triggers	Google Cloud Build	Auto-triggered production build and image compilation on git pushes
Analytics	GA4 Tracking	Google Analytics 4	Product usage tracking and conversion auditing
Quality Control	Jest + ts-jest	-	Automated unit testing suites for core parsing logic

---

📁 Repository Structure

lexguard-landing/
├── app/
│   ├── page.tsx                  # / — Landing Page (Marketing, features & Waitlist)
│   ├── layout.tsx                # Root layout, fonts, and Google Analytics 4
│   ├── globals.css               # Premium design tokens, animations, grids
│   ├── app/
│   │   ├── page.tsx              # /app — The interactive client dashboard tool
│   │   └── layout.tsx            # Full-height dashboard wrapper layout
│   └── api/
│       ├── analyze/
│       │   └── route.ts          # Server-side 3-Agent Gemini analysis pipeline
│       └── gemini/
│           └── route.ts          # Secure client proxy for manual Gemini queries
│
├── components/
│   ├── hero.tsx                  # Landing page blocks & waitlist module
│   ├── problem.tsx
│   ├── product-preview.tsx
│   └── tool/
│       ├── sidebar.tsx           # Dashboard sidebar, profile display & history loader
│       ├── user-onboarding.tsx   # Local storage user profile initializer
│       ├── upload-panel.tsx      # Text paste/drag-and-drop PDF upload component
│       ├── loading-screen.tsx    # Multi-agent simulated loading timers
│       ├── dashboard.tsx         # The main visual analysis dashboard UI
│       ├── clause-card.tsx       # Expandable clause cards with pushback copy utility
│       └── risk-summary.tsx      # Metrics header (Risk score, high/medium/low counts)
│
├── lib/
│   ├── types.ts                  # Shared strict TypeScript schemas
│   ├── gemini.ts                 # Secure server-side Gemini client
│   ├── firebase.ts               # Firebase App & Firestore Admin initializer
│   ├── history.ts                # Firestore CRUD (saveAnalysis, getUserHistory)
│   ├── risk-score.ts             # Mathematical risk-index grading algorithm
│   └── agents/
│       ├── parser.ts             # Agent 1 rules, prompt templates & validation
│       ├── analyzer.ts           # Agent 2 rules, prompt templates & validation
│       └── advocate.ts           # Agent 3 rules, prompt templates & validation
│
├── __tests__/                    # Automated Jest test suites
│   ├── agents-parser.test.ts     # Tests parser model boundaries
│   └── risk-score.test.ts        # Validates math engines for risk index scoring
│
├── Dockerfile                    # Multi-stage production Node 20 runner
├── next.config.ts                # Standalone compilation output configurator
└── package.json                  # Dependencies & task scripts

---

🚀 Getting Started Locally

Prerequisites

• Node.js v20.9.0 or higher (Strict requirement for Next.js 15/16)
• A Gemini API Key (Grab one instantly from Google AI Studio)

Step-by-Step Installation

1. Clone the repository:

   git clone https://github.com/MrCyberNaut/LexGuard.git
   cd LexGuard/lexguard-landing

2. Configure environment variables: Create a .env.local file inside the lexguard-landing/ directory:

   # lexguard-landing/.env.local
   GEMINI_API_KEY=YOUR_GEMINI_API_KEY_HERE
   
   # Optional - Required if using Firestore History Vault
   FIREBASE_SERVICE_ACCOUNT_KEY={"type":"service_account",...}

3. Install dependencies:

   npm install

4. Run the local development server:

   npm run dev

5. Open in your browser:

   • Landing & Waitlist: http://localhost:3000
   • Interactive App: http://localhost:3000/app

---

🧪 Automated Testing Suite

To ensure absolute precision when dealing with legal logic, LexGuard includes a comprehensive test suite to validate the integrity of our parsing engines and mathematical risk scorings.

To execute the test suites locally, run:

npm run test

---

🌐 Production Deployment Guide (GCP Cloud Run)

This repository is optimized to build inside a secure, lightweight Docker container and run on Google Cloud Run using continuous integration through Google Cloud Build.

1. Build Pipeline (Cloud Build)

Cloud Build is configured with a Git trigger that fires on every push to the main branch:

1. Pulls the latest code.
2. Compiles using a multi-stage Node 20 Alpine image.
3. Optimizes the output size using Next.js standalone file-tracing.
4. Pushes the compiled image to the GCP Artifact Registry.

2. Live Hosting (Cloud Run)

The container is deployed serverlessly in europe-west1 (Belgium):

• Port: 8080 (Mapped dynamically to Next.js host 3000)
• Scale: Automatically scales down to 0 instances when inactive (maximizing resource efficiency).
• Environment Variables: Configured securely in the GCP Console under the Variables tab (GEMINI_API_KEY).

3. Database Sync (Firestore)

When deployed on Cloud Run, the system automatically uses its native service account credentials to authenticate securely with Google Cloud Firestore. No local keys or passwords need to be committed to Git!

---

🎨 Premium UI & Design Tokens

LexGuard uses a bespoke Deep Dark UI theme ("Void Theme") constructed directly within vanilla CSS variables inside globals.css.

:root {
  --ink: #0e0e0e;         /* Pure dark, used in light overlays */
  --paper: #f5f4f0;       /* Warm-toned premium white */
  --void: #080808;        /* Pure void, tool background */
  --void-2: #111110;      /* Sidebar and secondary containers */
  --void-3: #1c1c1a;      /* Raised card surfaces */
  --accent: #5b48e8;      /* Electric purple primary indicator */
  
  --risk-red: #d93a2b;    /* High risk alert badge */
  --risk-amber: #c47a16;  /* Warning state */
  --risk-green: #1a7a4a;  /* Checked & safe state */
}

The interface is designed with:

• 🌌 Glassmorphism panels with subtle rgba border overlays.
• 📐 Blueprint layout grids matching high-fidelity engineering draft schematics.
• 📈 Micro-animations for transitions, agent load cycles, and hover states.

---

🛡️ License & Credit

Created under Problem Statement 01 by the LexGuard team for Google × Scaler PromptWars 2026. All code and methodologies represent modern engineering best-practices.