Built for how developers actually work
Alloy is a native desktop IDE — embedded Neovim editor, multi-session AI chat, plan mode, and a galaxy view of your work. Powered by the OpenAlloy engine across 13+ providers, with CLI, TUI and WhatsApp daemon also available.
Alloy — the AI coding IDE
A native desktop application for macOS, Windows, and Linux. Edit with an embedded Neovim, run multiple AI chat sessions side by side, track tasks in plan mode, and explore how your sessions connect in a 3D galaxy view.
- Embedded Neovim editor with side-by-side editing
- Multi-session chat, plan mode & extended thinking
- Galaxy view & workspace folder locking
Native desktop IDE with embedded Neovim, multi-session AI chat, plan mode, and a galaxy view — all backed by the OpenAlloy engine.
Rendered by ingot
Alloy's UI isn't a browser wrapper. It's drawn by ingot — a pure-Odin, immediate-mode engine built directly on WebGPU. A single source targets native GPUs (Metal, D3D12, Vulkan) and the browser (WASM + WebGPU), with no external dependencies to vendor.
- GPU-accelerated 2D rendering with glyph-atlas text
- Embedded terminal stack & native custom window chrome
- Adaptive frame pacing, crisp at any resolution × DPI
Multi-Model Orchestration
Switch between LLM providers mid-session without losing context. Use Claude for complex reasoning, GPT for code generation, Gemini for multimodal tasks - all from the same interface.
- Anthropic, OpenAI, Google, Groq, Mistral, Cohere
- OpenRouter, AWS Bedrock, Azure OpenAI
- Local models via Ollama & LMStudio
- GitHub Copilot integration
Team Setup Tokens
Onboard developers in seconds. Generate time-limited setup tokens that automatically configure provider access, team settings, and telemetry for new team members.
- Time-limited tokens with configurable expiry
- Automatic provider & permission configuration
- Seat management and usage caps
Webhook Usage Monitoring
Every API call generates an HMAC-signed webhook event with token counts, cost estimates, model info, and session metadata. Pipe it to your observability stack or use our built-in dashboard.
- HMAC-SHA256 signed payloads
- Per-user, per-model, per-session breakdowns
- Real-time cost tracking and alerts
WhatsApp Daemon
Turn any WhatsApp chat into an AI-powered workspace. Also available beyond the desktop app: run the OpenAlloy engine as a background daemon that listens for messages, processes prompts, and responds — all from your phone. Supports text, voice notes, images, and documents.
- Scan a QR code to pair — no WhatsApp Business API needed
- Persistent sessions per chat (DMs and groups)
- Send images, voice notes, PDFs — the agent sees everything
- Allowlist phone numbers for security
And much more
LSP Integration
Language Server Protocol support gives the agent deep code understanding — go-to-definition, references, and diagnostics inform every edit.
MCP Extensibility
Model Context Protocol support lets you extend Alloy with custom tools, data sources, and capabilities.
Works Everywhere
Native binaries for macOS (Intel & Apple Silicon), Linux, Windows, Android, and BSD. No runtime required.
Terminal TUI
Also available: a beautiful terminal interface built on Charm's Crush. Full mouse support, markdown rendering, and syntax highlighting.
Terminal Agent
Autonomous agent mode that reads files, runs commands, and writes code. Full tool-use support with approval workflows.