Aurevoy

Tell it what you want, not how to do it.
Your first AI companion that actually gets things done.

Aurevoy is a personal AI agent that runs on your computer. It's more than a chat interface —
you give it a goal, and it figures out the rest: researching, writing files, searching the web, and working through tasks step by step.

Like having someone sitting next to you who just says "I'll take care of it."

What It Can Do

Everyday Tasks, Just Ask

"Convert all the PNGs in my downloads folder to WebP"
"Research React 19 breaking changes and write a migration checklist"
"Read through this folder and turn the code into a README"

Aurevoy breaks down the task and gets to work. You don't need to explain how.

Write, Edit, and Explore Code

"Add dark mode to this project"
"Find the date-formatting utility in src/utils and check for bugs"
"This code is too slow — help me find the bottleneck"

It understands your project structure, makes precise edits, and flags issues. If something goes wrong, you can always undo — edit any earlier message, branch the conversation, or compress context to start fresh.

Research & Writing

"Survey the top React state management libraries — compare pros and cons"
"Translate this article into English, keeping technical terms accurate"
"Turn these meeting notes into a weekly report"

It searches the web for up-to-date information, fetches pages for details, and synthesizes everything into the format you need.

See and Understand Images

Drag and drop images or files into the conversation. Aurevoy switches to a vision model when images are present, so it can read screenshots, analyze diagrams, and answer questions about what it sees.

Learns Your Preferences

Tell it how you work — "I always use pnpm," "I prefer functional style" — and it remembers. Next time it follows your conventions without being reminded.

How It Works

You give it a goal → Aurevoy scouts your workspace to understand the context → builds a plan → executes step by step:

Need to read a file? It opens and inspects it.
Need to look something up? It searches the web for current information.
Need to write code? It makes precise edits without touching unrelated parts.
Need your input? It pauses, asks, and waits for your go-ahead before proceeding.

You see every step. Dislike a direction? Interrupt anytime, revise, and it picks back up. You can also rewind — edit any earlier message and regenerate from that point, or branch into a new thread to explore a different direction.

Your Data, Your Rules

Everything runs locally. Your data never leaves your machine.
You bring your own API key and choose whichever model you want.
The code is open source — you can always see exactly what it's doing.

Get Started

1. Download

Grab the latest DMG from Releases and drag it into Applications.

2. Configure a Model

Open Aurevoy and fill in your model details under Settings:

Provider: OpenAI (or any compatible API)
Base URL: https://api.openai.com/v1
Model: gpt-4o-mini
API Key: sk-xxxx

Works with OpenAI, DeepSeek, Ollama, and any OpenAI-compatible endpoint. Use your own key — no extra fees.

3. Start a Conversation

Tell Aurevoy what you need in the input box. Talk to it like a person.

/  Browse available skills (web-search, browser, and more)

Contributing

Aurevoy is open source. Contributions are welcome.

# Prerequisites: Node.js >= 20, Rust (stable), macOS
git clone https://github.com/nullskymc/Aurevoy.git
cd Aurevoy
npm install
npm run dev          # start dev mode
npm run typecheck    # type check
npm run build        # production build

More details in AGENTS.md and docs/.

Tech Stack

Layer	Technology
Desktop App	Tauri 2 + React + TypeScript
Agent Engine	Node.js + Fastify + SQLite
Transport	Local HTTP + SSE streaming
Tool Extensions	Built-in tools + MCP protocol (connect external tool servers)
Skill System	Slash-command-activated capabilities, custom skills supported

See docs/ARCHITECTURE.md for details.

What's Included

Autonomous planning & parallel tool execution
File read/write/search/copy/move/delete, precision diff edits
Web search & page fetching with SSRF protection
Long-term memory (learns your preferences over time)
Multi-turn conversations, edit & retry, branching threads, context compression
Multimodal: image/file drag-and-drop, vision model support
Project workspaces for organizing conversations
Skill system (slash commands, custom skills)
macOS native desktop app

Coming next: knowledge base / RAG, agent evaluation framework, release polish. See ROADMAP.

MIT