Open source · MIT · macOS / Linux / Windows
Your machine.
Your agent. Your story.
onfable is an open-source autonomous AI agent that lives in your terminal — it runs commands, edits files, browses the web, and remembers you. Bring any model: Claude, OpenAI, OpenRouter, or your own endpoint.
$ curl -fsSL https://onfable.xyz/install.sh | shSee it in action
One sentence in. Real work out — with your approval on every command.
Installed in one line
Node.js 20+ is the only prerequisite. Your API key stays on your machine.
$ curl -fsSL https://onfable.xyz/install.sh | shThen run onfable setup — pick a provider, paste your key, done.
An agent, not a chatbot
Chatbots tell you how. onfable just does it.
Runs 100% locally
onfable lives on your machine — macOS, Linux, or Windows. Your API key, your files, your data: everything stays in ~/.onfable, nothing passes through our servers.
Works with any model
Claude by default, or OpenAI, OpenRouter, and any OpenAI-compatible endpoint. Switch providers with one command, override the model per session.
Persistent memory
It remembers you. Preferences, project context, how you like things done — saved to a plain markdown file you can read and edit yourself.
Power with guardrails
Full system access — shell, files, the lot — but every command and file write asks for your approval first. Trust it? Run with --yolo.
Web-aware
Built-in web search and page fetching. It can look things up, read documentation, and pull real data into your tasks — no extra API key needed.
Open source & hackable
MIT licensed, one lean TypeScript codebase. Add a tool in ~40 lines. Read every prompt it sends. Fork it and make it yours.
Up and running in a minute
Install
One line in your terminal. macOS, Linux, or Windows.
curl -fsSL onfable.xyz/install.sh | shSetup
Pick your provider, paste your API key. Saved locally, chmod 600.
onfable setupAsk anything
It plans, runs the tools, and reports back — approving with you along the way.
onfable run "clean up my git branches"Meet it where you are
The terminal is home base. Messaging channels are on the roadmap.
Interactive REPL + one-shot runs
Your agent in your pocket
Agent for your server
Questions, answered
Is it free?+
Yes — onfable is MIT-licensed open source and free forever. You only pay your AI provider for the tokens you use (bring your own API key).
Which models can I use?+
Anthropic Claude (recommended), OpenAI, anything on OpenRouter, or any OpenAI-compatible endpoint — including local servers like Ollama or LM Studio that expose one. Switch anytime with `onfable setup`.
Is it safe to let an AI run commands on my machine?+
onfable is built around approval-first execution: every shell command, file write, and file edit shows you exactly what it wants to do and waits for your yes. Reads don't need approval. When you trust a workflow, `--yolo` skips the prompts.
Where is my data stored?+
Only on your machine. Your config (API key, chmod 600), memory, and session history live in ~/.onfable/. Conversations go directly from your machine to your chosen AI provider — there is no onfable server in between.
How is this different from a chatbot?+
A chatbot tells you the commands to run. onfable runs them — it lists your files, edits them, searches the web, retries when something fails, and remembers your preferences for next time. It's an agent with hands, not just a voice.