Choeng-Rayu/rayu-cli

🎥 Watch It In Action

🎬 RAYU CLI in action — 4:42 demo showing multi-provider AI coding agent capabilities

📥 Download MP4 (13MB) · 📥 Download WebM (31MB) · 📖 Documentation

The ultimate standard for terminal-based AI development. Rayu is built from the ground up to outperform every other CLI agent on the market. It works flawlessly, runs faster, and respects your privacy completely.

---

⚡ Why Rayu Outperforms the Rest

Feature	Rayu	OpenCode	KiloCode	Claude Code	Codex
Response Speed	⚡ Near-instant (<500ms)	~3-5s	~5-10s	~5-8s	~3-6s
P2P Collaboration	✅ Native Peer-to-Peer	❌ No	❌ No	❌ No	❌ No
Privacy / No Training Data	✅ Zero Data Retention	⚠️ Cloud-dependent	⚠️ Cloud-dependent	❌ Trains on your code	❌ Trains on your code
Multi-Provider BYOK	✅ Anthropic, OpenAI, DeepSeek, etc.	❌ Limited	❌ Single	❌ Anthropic only	❌ OpenAI only
Agent Swarms	✅ Built-in & Parallel	❌ No	❌ No	❌ No	❌ No
Offline Capabilities	✅ Private local model support	❌ No	❌ No	❌ No	❌ No

---

🚀 Key Advantages

1. Blazing Fast Speed

Rayu is built on high-performance streaming architectures, eliminating the typical latency bottlenecks of traditional CLI tools. Our terminal UI utilizes a custom React/Ink renderer with zero-GC cell buffers, and our Go gateway ensures sub-millisecond routing overhead.

• Time-to-first-token: Under 500ms for cached sessions and 1-2s for cold boots.
• No more waiting on sluggish web bridges or processing screens.

2. Peer-to-Peer (P2P) Collaboration

Rayu is the only CLI agent with native P2P capabilities. Collaborate directly with team members in real-time without sending intermediate payloads to external cloud storage. This is ideal for secure enterprise environments, air-gapped systems, and lag-free cooperative terminal sessions.

3. Absolute Privacy: No Training Data

While other platforms use your inputs and proprietary source files to train their models, Rayu never trains on your code.

• BYOK (Bring Your Own Key): Connect directly to your preferred API keys.
• Hosted Gateway: Completely transparent proxy pipelines with zero-data-retention guarantees.

4. Flawless Execution & Freedom of Choice

Rayu works perfectly right out of the box. Switch models instantly mid-session with /model or run parallel diagnostics on your codebase across multiple LLM providers (Anthropic, DeepSeek, Google Gemini, OpenAI, Kimi, and locally hosted models via Ollama/LM Studio).

---

💻 Quick Start

One command, any machine — no Node, npm or sudo required:

# macOS / Linux
curl -fsSL https://rayucode.com/install | bash

# Windows (PowerShell)
irm https://rayucode.com/install.ps1 | iex

Then:

rayu

Already have Node ≥ 18 and would rather use npm?

npm install -g @rayu-dev/rayu-cli

Or run instantly without installing:

npx @rayu-dev/rayu-cli

On first launch, Rayu will guide you through connecting a provider. Choose BYOK (your own API key) or a Rayu-hosted connection.

---

🌐 Monorepo Services

Rayu consists of four independent services. They work together or as individual standalone configurations.

Directory	Stack	Role
rayu/	TypeScript + Bun + Ink	The CLI — AI coding agent itself
rayu-backend/	NestJS + Prisma + MySQL	Accounts API — auth, plans, billing
rayu-gateway/	Go + chi + Redis	AI Gateway — streaming proxy, rate limiting
rayu-web/	Next.js 15 + NextAuth	Website — marketing site + user dashboard
deploy/	Docker Compose + Caddy	Production — single-VPS deployment stack

---

🛠️ CLI Quick Commands

cd rayu && bun install && bun run dev       # Run CLI from source
cd rayu-backend && npm install && npm run start:dev   # Launch accounts backend
cd rayu-gateway && go run ./cmd/gateway      # Run gateway routing proxy
cd rayu-web && npm install && npm run dev    # Spin up web dashboard

---

🤝 Contributing

We welcome contributions! Please see our Contributing Guide for details on issues, code formatting, and pull requests.

---

🌐 rayucode.com · Issues · NPM