🗂️ BigRack.dev

MCP Server for Task & Context Management in AI Development

⚠️ BETA PREVIEW: This project is currently in public beta. Interfaces and database schemas may change. Use with caution in production environments.

BigRack is a Model Context Protocol (MCP) server that extends AI development tools like Claude Code and Cursor with advanced task/project management and persistent context capabilities.

🚀 Setup

1. Build the MCP Package

# From the repository root
npm run build:mcp

2. Link the Package Globally

# Link for local development
npm link

Or using the npm script:

npm run link:mcp

3. Initialize BigRack Globally

# Run once per machine (initializes ~/.bigrack/)
bigrack init

This sets up the global BigRack environment on your machine.

4. Configure Your AI Assistant

For Claude Desktop:

bigrack setup-claude

Then restart Claude Desktop.

For Cursor:

bigrack setup-cursor

📖 What's Next?

After setup, you can:

• Initialize a project: In your project directory, use Claude/Cursor to call bigrack_create_repo - this creates bigrack.json and links your project to BigRack
• Store business context: Ask Claude/Cursor to store business rules, glossary terms, or architecture patterns
• Create projects: Create a new project (Pallet) for a feature or bugfix
• Decompose features: Break down complex features into atomic tasks with dependencies
• Query context: Use semantic search to find relevant business rules and patterns

🎯 The Problem

AI assistants like Claude Code are powerful but face major limitations on complex projects:

• Progressive context loss: Details mentioned early in conversations are forgotten
• Insufficient planning: Difficulty managing complex tasks with dependencies
• Premature simplification: Missing edge cases and important business validations
• Forgotten business rules: Critical constraints disappear during development

💡 The Solution

BigRack provides a structured, persistent context that AI can always consult through the Model Context Protocol. Think of it as giving your AI assistant a long-term memory for your project's business rules, architecture decisions, and task dependencies.

✨ Key Features

📡 MCP Protocol

Standard protocol to extend AI assistants with custom tools. Works with:

• Claude Code (Anthropic)
• Cursor (Anysphere)
• Any MCP-compatible AI assistant

📦 Task Decomposition

• Break down complex features into atomic tasks
• Automatic dependency graph (DAG) creation
• Intelligent task recommendations based on priority and blockers
• Track progress with pending, in-progress, completed, blocked statuses

🧠 Context Management

• Store business rules, glossary terms, architecture patterns, and team conventions
• Semantic search powered by vector embeddings (Xenova/all-MiniLM-L6-v2, 384 dimensions)
• RAG (Retrieval-Augmented Generation) for intelligent context retrieval
• Query context with natural language: "What are the authentication requirements?"

🏠 Local-First Architecture

• 100% local - all data stored on your machine with SQLite
• Embedded vector search - no external APIs or services required
• Zero cloud dependency - works completely offline
• Privacy-first - your code never leaves your machine

📚 Core Concepts

Rack (Repository)

A Rack represents a business domain or codebase. It's the main container for:

• Business rules
• Glossary terms
• Architecture patterns
• Team conventions
• One Rack per project directory (identified by bigrack.json)

Project

A Project represents a specific work unit (feature, bugfix, refactor):

• Automatically inherits context from parent Rack
• Can have project-specific context
• Contains tasks with dependencies
• Linked to Git branches

Task

An atomic task with:

• Title, description, status, priority, type
• Dependencies (other tasks that must be completed first)
• Validation criteria
• Estimated time
• Git branch tracking

Vector Search & RAG

• All business context is automatically embedded using Xenova/all-MiniLM-L6-v2 (~22.6 MB)
• 384-dimensional vector embeddings
• Natural language queries return the most relevant context
• Runs 100% locally with no external API calls

🛠️ CLI Commands

Global Setup

bigrack init                    # Initialize BigRack globally (once per machine)
bigrack status                  # Show current repository status

Project Management

bigrack projects create --name "Feature X" --type feature
bigrack projects list

Task Management

bigrack ticket list                           # List all tasks
bigrack ticket list --status pending          # Filter by status
bigrack ticket list --priority critical       # Filter by priority
bigrack ticket get <ID>                       # Show task details

Context Management

bigrack context add --type business_rule --name "Rule" --description "..."
bigrack context query "authentication requirements"
bigrack context list

Configuration

bigrack config show             # Show current configuration
bigrack config get <key>        # Get config value
bigrack config set <key> <value> # Set config value

🏗️ Architecture

bigrack.dev/
├── apps/
│   ├── website/              # Marketing website (Next.js 14)
│   ├── api/                  # Backend API (future)
│   └── dashboard/            # Web dashboard (future)
├── packages/
│   ├── mcp/                  # @bigrack/mcp - MCP server (THIS IS THE CORE)
│   │   ├── src/
│   │   │   ├── mcp/          # MCP server & tools
│   │   │   ├── cli/          # CLI commands
│   │   │   ├── storage/      # SQLite + Prisma
│   │   │   ├── embeddings/   # Vector embeddings
│   │   │   └── logger/       # Pino logger
│   │   └── prisma/           # Database schema
│   ├── shared/               # Shared types & utilities
│   └── config/               # Shared configs
└── LICENSE                   # Apache License 2.0

📦 Technology Stack

MCP Server (Local)

• Runtime: Node.js 20+, TypeScript 5.3+
• Database: SQLite via Prisma ORM
• Vector Search: Xenova/transformers.js (local embeddings)
• Model: all-MiniLM-L6-v2 (384 dimensions, ~22.6 MB)
• CLI: Commander, Inquirer, Chalk, Ora
• Logging: Pino

Website

• Framework: Next.js 14 (App Router)
• Styling: Tailwind CSS + shadcn/ui
• Deployment: Vercel

🤝 Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.

Development Setup

# Clone repository
git clone https://github.com/baptiste-mnh/bigrack.dev.git
cd bigrack

# Install dependencies
npm install

# Build all packages
npm run build

# Build MCP package
npm run build:mcp

# Link MCP package for local development
npm run link:mcp

# Run in development mode with auto-rebuild
npm run dev:mcp

Project Structure

• Monorepo managed with Turborepo
• Workspaces: apps/* and packages/*
• MCP package: packages/mcp/

Commit Conventions

Use Conventional Commits:

feat(mcp): add ticket list command
fix(storage): resolve dependency resolution bug
docs(readme): update installation instructions

📄 License

BigRack is licensed under the Apache License 2.0.

This means you can:

• ✅ Use commercially
• ✅ Modify and distribute
• ✅ Use privately
• ✅ Sublicense

With protections for:

• 🛡️ Patents (patent grant included)
• 🛡️ Trademarks (you can't use "BigRack" name without permission)

See LICENSE for full details.

⚠️ Disclaimer

This software is provided as-is, without any warranty.

All data is stored locally on your device.

See the Apache 2.0 license for more details.

🌟 Why BigRack?

For Solo Developers:

• Maintain context across long development sessions
• Never lose track of business rules and constraints
• Plan complex features with confidence

For Teams:

• Share business context across the team
• Consistent AI-assisted development
• Track progress and dependencies

For Enterprises:

• Manage complex projects with strict business rules
• Compliance and audit trails
• Self-hosted for data privacy

🔗 Links

• Website: bigrack.dev
• Documentation: bigrack.dev/docs
• GitHub: github.com/baptiste-mnh/bigrack.dev
• npm Package: @bigrack/mcp
• Issues: github.com/baptiste-mnh/bigrack.dev/issues
• Discussions: github.com/baptiste-mnh/bigrack.dev/discussions

---

Built with ❤️ by @baptiste-mnh

Empowering developers to build complex projects with AI assistance