Æther Shell

The world's first multi-agent shell with typed functional pipelines and multi-modal AI.
Built in Rust for safety and performance, featuring revolutionary AI protocols found nowhere else.

Quick Start • Features • Examples • TUI Guide • Docs • Contributing

🚀 Quick Start

# Install from source
git clone https://github.com/nervosys/AetherShell && cd AetherShell
cargo install --path . --bin ae

# Launch interactive TUI (recommended)
ae --tui

# Or classic REPL
ae

# Typed pipelines — not text streams!
[1, 2, 3, 4, 5] | map(fn(x) => x * 2) | sum()
# => 30

# AI query
ai("Explain quantum computing in simple terms")

# AI with vision
ai("Describe this image", {images: ["photo.jpg"]})

# AI agent with tool access
agent("Find all TODO comments in src/", ["ls", "cat", "grep"])

# 130+ MCP tools
mcp_tools() | len()  # => 130

📝 Note: Set OPENAI_API_KEY for AI features: export OPENAI_API_KEY="sk-..."

✨ Features

🎯 What Makes AetherShell Unique?

AetherShell is the only shell combining these capabilities:

📖 Examples

AI Agents with Tool Access

# Deploy an AI agent that can use shell tools
agent("Analyze the project structure and find large files", ["ls", "cat", "wc"])

# Agent with configuration
agent({
  goal: "Find security issues in the codebase",
  tools: ["grep", "cat", "ls"],
  max_steps: 10,
  dry_run: true  # Preview actions first
})

Multi-Modal AI

# Analyze images
ai("What's in this screenshot?", {images: ["screenshot.png"]})

# Process audio
ai("Transcribe and summarize this meeting", {audio: ["meeting.mp3"]})

# Video analysis
ai("Extract the key steps from this tutorial", {video: ["tutorial.mp4"]})

Typed Functional Pipelines

# Structured data processing — not text parsing!
ls("./src")
  | where(fn(f) => f.ext == ".rs" && f.size > 1000)
  | map(fn(f) => {name: f.name, kb: f.size / 1024})
  | sort_by(fn(f) => f.kb, "desc")
  | take(5)

# Statistical operations
[1, 2, 3, 4, 5] | sum()      # => 15
[10, 20, 30] | avg()         # => 20.0
[1, 2, 1, 3] | unique()      # => [1, 2, 3]
{a: 1, b: 2} | values()      # => [1, 2]

MCP Tools (Model Context Protocol)

# 130 tools across 27 categories
let tools = mcp_tools()
print(len(tools))  # => 130

# Filter by category
mcp_tools({category: "development"})     # git, cargo, npm, etc.
mcp_tools({category: "machinelearning"}) # ollama, tensorboard, etc.
mcp_tools({category: "kubernetes"})      # kubectl, helm, k9s, etc.

# Execute tools via MCP
mcp_call("git", {command: "status"})

Neural Networks & Evolution

# Create a neural network
let brain = nn_create("agent", [4, 8, 2])

# Evolutionary optimization
let pop = population(100, {genome_size: 10})
let evolved = evolve(pop, fitness_fn, {generations: 50})

# Reinforcement learning
let agent = rl_agent("learner", 16, 4)

🎮 TUI Interface

Launch the beautiful terminal UI with ae --tui:

Keyboard shortcuts:

• Tab — Switch tabs
• Enter — Send message / activate
• Space — Select media files
• q — Quit
• Ctrl+C — Force quit

📖 Full guide: docs/TUI_GUIDE.md

📦 Installation

From Source (Recommended)

git clone https://github.com/nervosys/AetherShell
cd AetherShell
cargo install --path . --bin ae

From Crates.io

cargo install aether_shell

VS Code Extension

Get syntax highlighting, snippets, and integrated REPL:

cd editors/vscode
npm install && npm run compile
# Press F5 to test

⚙️ Configuration

Environment Variables

# AI Provider (required for AI features)
export OPENAI_API_KEY="sk-..."

# Agent permissions
export AGENT_ALLOW_CMDS="ls,git,curl,python"

# Alternative AI backend
export AETHER_AI="ollama"  # or "openai"

Secure Key Storage

# Store keys in OS credential manager (recommended)
ae keys store openai sk-your-key-here

# View stored keys (masked)
ae keys list

📚 Documentation

Example Scripts

🧪 Testing

# Run all tests (90 library tests)
cargo test --lib

# Run specific test suites
cargo test --test eval          # Evaluator tests
cargo test --test mcp           # MCP protocol tests
cargo test --test tui_          # TUI tests

Test coverage: 140+ tests covering core functionality, MCP protocol, TUI, AI backends, neural networks, and OS tools.

🛣️ Roadmap

✅ Completed (January 2026)

• Neural network primitives & evolutionary algorithms
• 130+ MCP tools with protocol compliance
• Multi-modal AI (images, audio, video)
• Reinforcement learning (Q-Learning, DQN, Actor-Critic)
• Distributed agent swarms
• VS Code extension

🔜 Coming Soon

• Plugin system for custom backends
• WASM support (browser-based shell)
• Advanced video streaming
• Package management & imports

🤝 Contributing

We welcome contributions! See our development setup:

git clone https://github.com/nervosys/AetherShell
cd AetherShell
cargo build
cargo test --lib

1. Fork the repository
2. Create a feature branch
3. Add tests for new functionality
4. Submit a pull request

📜 License

Licensed under the Apache License 2.0.

Ready to experience the future of shell interaction?

ae --tui

⭐ Star us on GitHub • 🐛 Report Issues • 💬 Discussions