GitHub Docs

VibeCody

AI-powered developer toolchain built in Rust.

VibeCody is a monorepo that provides a full family of AI-assisted development surfaces:

  • VibeCLI — terminal-first AI coding assistant with a rich TUI + REPL
  • VibeUI — AI-powered desktop code editor (Tauri + Monaco, 293+ panels)
  • VibeCLI App — dedicated desktop chat companion (Tauri)
  • VibeMobile — Flutter companion for iOS, Android, macOS, Linux, Windows, Web
  • VibeWatch — native Apple Watch (SwiftUI, watchOS 10+) and Wear OS (Compose, Wear OS 3+) clients
  • Zero-config connectivity — mDNS LAN, Tailscale Funnel, ngrok auto-detect; devices race all reachable paths

All surfaces are backed by a shared set of Rust crates for AI providers, text editing, Git, LSP, and extensions.

Page Description
Easy Setup Deploy VibeCody anywhere — cloud, desktop, or Raspberry Pi
Use Cases 80+ things you can do with VibeCody
Deployment Guides Step-by-step guides for 12 platforms (AWS, GCP, Azure, Pi, etc.)
Quickstart Zero to productive in 5 minutes
Tutorials Step-by-step guides for common workflows
VibeCLI Reference CLI commands, TUI usage, and configuration
VibeUI Reference Desktop editor features and setup
Design System Token-based UI system — colors, spacing, typography, components
Configuration Guide All configuration options for providers and UI
Memory Guide All memory layers — auto-recording, cognitive store, verbatim drawers, benchmarking
Memory Architecture Detailed architecture: five stores, Context Assembler, Recap/Resume, storage security
Architecture Crate structure, data flow, and design decisions
Roadmap Competitive landscape, phase-level history (1–39), and positioning
Plugin Development Build plugins, skills, hooks, WASM extensions, and MCP integrations
VibeMobile Mobile companion (Flutter) — pairing, Handoff, remote chat, sync
VibeWatch — watchOS Apple Watch native client — pair, view transcripts, dictate reply
VibeWatch — Wear OS Wear OS native client — same feature set, Keystore/StrongBox attestation
Connectivity mDNS, Tailscale Funnel, ngrok — zero-config device discovery
Watch Integration Full architecture: P-256 pairing, /watch/* routes, sync model
Competitive Analysis Feature comparison with other AI coding tools
Fit-Gap Analysis Consolidated gap catalogue — 142 gaps tracked across 8 iterations and 5 deep-dives
Whitepapers In-depth comparisons: VibeCody vs OpenClaw, PicoClaw, NemoClaw, and 12+ alternatives
Development Guide Build, test, debug, and code organization for contributors
Security Security model, SSRF/path-traversal prevention, command blocklists
Release Notes What’s new in v0.4.0 — downloads, upgrade guide
Contributing How to build, test, and contribute

Key Features

Multi-Provider AI (23 Providers)

Both VibeCLI and VibeUI share the vibe-ai provider abstraction:

Category Providers
Local Ollama, LocalEdit
Cloud Claude, OpenAI, Gemini, Grok, Groq, Mistral, Cerebras, DeepSeek, Zhipu, MiniMax
Platform OpenRouter, Azure OpenAI, Bedrock, Copilot, Vercel AI
Inference Perplexity, Together AI, Fireworks AI, SambaNova
Meta Failover (automatic provider fallback)

All providers support streaming. Local providers require no API key.

VibeCLI Highlights

  • Rich TUI powered by Ratatui with REPL mode (readline history, tab completion)
  • 556 skill files across 25+ categories (106+ REPL commands)
  • Voice input via Groq Whisper (--voice flag)
  • Tailscale pairing with QR code sharing and mDNS discovery
  • 18 gateway platforms (Telegram, Discord, Slack, Signal, Matrix, Teams, IRC, Twitch, and more)
  • Red team security pipeline and compliance reporting
  • Workflow orchestration with 8-stage Code Complete pipeline
  • MCP (Model Context Protocol) server — 51 tools including email, calendar, tasks, Notion, Jira, Home Assistant
  • Productivity integrations: Gmail/Outlook, Google/Outlook Calendar, Todoist, Notion, Jira, Home Assistant
  • 12-platform deployment: AWS, GCP, Azure, Oracle Cloud, DigitalOcean, Linode, macOS, Linux, Windows, Raspberry Pi 3/4/5
  • Session persistence with /sessions and /resume commands
  • Container sandbox (Docker, Podman, OpenSandbox) with unified runtime trait
  • Git-aware context injection, multi-file diff view with syntax highlighting
  • AI-assisted code apply with interactive confirmation and approval gate
  • HTTP daemon mode (vibecli --serve --port 7878)

VibeUI Highlights

  • Monaco Editor integration (same engine as VS Code)
  • 196+ AI panel tabs (Chat, Agent, Counsel, Tests, Docker, K8s, Profiler, Design Canvas, and many more)
  • CSS variable theming across all panels with dark/light toggle
  • Multiplayer CRDT collaboration (real-time co-editing)
  • Agent teams with inter-agent messaging bus
  • CI review bot (GitHub App integration)
  • Marketplace for sharing extensions and skills
  • Visual testing via Chrome DevTools Protocol
  • Container sandbox management (Docker, Podman, OpenSandbox)
  • Deploy to 6 targets from the editor
  • Rope-based text buffer, async file I/O with file-watching
  • Full Git panel, integrated terminal (PTY), LSP client, WASM extension system

VibeMobile (Flutter Companion)

  • QR code pairing with VibeCLI/VibeUI instances
  • Remote AI chat from iOS, Android, macOS, Linux, Windows, Web
  • Machine management with health monitoring
  • Session browser and push notifications

Getting Started in 60 Seconds

# Clone
git clone https://github.com/TuringWorks/vibecody.git
cd vibecody

# Build VibeCLI
cargo build --release -p vibecli

# Launch the TUI (requires Ollama running locally)
./target/release/vibecli --tui

# Or launch VibeUI
cd vibeui && npm install && npm run tauri dev

See the Configuration Guide to set up cloud providers.