Atlas UI 3

A modern LLM chat interface with MCP (Model Context Protocol) integration.

Note

This App is still under development. Not all features work well.

Read this first (for contributors)

For the most current, concise guidance on how to work in this repository, read:

• CLAUDE.md — architecture, workflows, and conventions verified against code
• .github/copilot-instructions.md — a compact “AI agent guide” for getting productive fast

These two docs are kept up-to-date more frequently than long-form docs. Start there before exploring the wider docs or codebase.

Features

• Multi-LLM Support: OpenAI GPT, Anthropic Claude, Google Gemini
• MCP Integration: Connect to multiple MCP servers for tools and data sources
• Real-time Communication: WebSocket-based chat interface
• Custom UI: MCP servers can modify the UI with custom HTML
• Authorization: Group-based access control for MCP servers
• Modern Stack: React frontend, FastAPI backend, Docker support

Quick Start

Docker (Recommended)

docker build -t atlas-ui-3 .
docker run -p 8000:8000 atlas-ui-3

Open http://localhost:8000

Local Development

Important: This project uses uv as the Python package manager.

# Install uv if needed
curl -LsSf https://astral.sh/uv/install.sh | sh

# Setup environment
uv venv && source .venv/bin/activate  # Windows: .venv\Scripts\activate
uv pip install -r requirements.txt

# Configure
cp .env.example .env  # Edit with your API keys

# Build frontend
cd frontend && npm install && npm run build

# Start backend
cd ../backend && python main.py

# OR the quickest way to start is to use the agent_start.sh

bash agent_start.sh

Key Technologies

• Backend: FastAPI + WebSockets
• Frontend: React + Vite + Tailwind CSS
• Python Package Manager: uv (not pip!)
• Configuration: Pydantic with type safety
• Containerization: Docker

Important Notes

• Use uv for Python package management, not pip or conda
• Don't use uvicorn --reload - causes problems in development
• Use npm run build instead of npm run dev for frontend development
• File limit: Maximum 400 lines per file for maintainability
• Container Environment: Use Fedora latest for Docker containers (GitHub Actions uses Ubuntu runners)
• Mock S3: The included S3 mock (mocks/s3-mock/) is for development/testing only and must NEVER be used in production due to lack of authentication, encryption, and other critical security features.

License

Copyright 2025 National Technology & Engineering Solutions of Sandia, LLC (NTESS). Under the terms of Contract DE-NA0003525 with NTESS, the U.S. Government retains certain rights in this software

MIT License