Meet FangBot.
AI agents that run on your Mac. Give them a job in chat, they use your tools and channels, and finish it end to end — asking only when it matters.
Message agents like teammates
Streaming answers, live tool calls, attachments and slash commands. Give tasks the way you would a colleague — agents keep context on how you work and come back only when your approval is needed.
Every agent gets a seat in the room
A 3D room on WebGL where each agent is a character with its task count on the nameplate — busy agents glow while they think. Click a character to open its chat. Prefer a plain list? One toggle.
The agent runs on your machine
Local mode is the default. FangBot runs an OpenClaw gateway as a background service on your Mac — no sign-in, keys stay on disk with owner-only permissions, and quitting the app never stops the agent.
Teach a task once
Show an agent a workflow once and it saves it as a skill — a SKILL.md file in the workspace; writing the file is the deployment. Author skills in Agent Studio with templates and live validation, then run them on a schedule.
Agents get smarter over time
Named agents keep memory, files and preferences across sessions — personality and identity are files you can open and edit right in the chat. Context compounds instead of resetting.
Answer where customers already are
Connect WhatsApp by QR, Telegram and Discord by bot token, Slack over Socket Mode — and the agent answers comments on your Facebook and Instagram pages through the browser on this machine.
A real Mac citizen
Not a web page in a frame. Menu bar, global shortcuts, floating panels, Keychain, auto-update — the parts only a native app can do.
A floating panel over every app — full-screen included.
Jump to rooms, chats, skills and tasks; restart the agent.
Live agent state in the tray, one click to chat.
Windows share one gateway connection.
Tokens in the Keychain; an Ed25519 key signs every handshake.
Downloads in the background — never relaunches mid-answer.
Give each agent a job
Start with one. Add more when the work grows.
Store Operations
Watches orders, flags fulfillment issues, and drafts the daily ops report before you open your laptop.
Customer Support
Answers tickets and comments in your brand voice, escalates edge cases, and leaves a review list for you to approve.
Ad Performance
Pulls spend and ROAS across your Meta campaigns, spots fatigue, and drafts creative briefs for the next test.
Review Analysis
Reads every new review, clusters the complaints, and tells you what to fix first — with receipts.
Content & SEO
Researches keywords, drafts product copy in your voice, and publishes to your CMS on approval.
Chief of Staff
The catch-all. Anything you keep doing twice a week, hand it over and check the approvals queue.
Get FangBot for macOS
Free to download. Local mode by default — no account, no cloud.
All releases · Changelog · macOS 13+ · signed & notarized
Questions
How is FangBot different from an AI chatbot?
Chatbots draft text. FangBot agents have a runtime on your machine — a browser, files, scheduled tasks, messaging channels — so they finish jobs in your real tools, end to end, and only surface when something needs your approval.
Does it work without an account?
Yes. Local mode is the default: no sign-in, no cloud. The agent runs as a background service on your Mac and your model keys stay on the machine. FangBot Cloud is optional.
Which platforms are supported?
FangBot Desktop is available today for macOS — both Apple Silicon and Intel. Windows and mobile are on the roadmap.
Which models can it run?
Bring your own model. If Ollama or LM Studio is already running on your Mac, FangBot finds it — no key needed. Otherwise paste a key for Anthropic, OpenAI, Google Gemini, OpenRouter, Groq and more.
How do agents use my API keys?
Keys are written to ~/.fangbot/.env on your Mac with owner-only permissions, and injected into the local runtime at launch. They never sit inside the app bundle and are never sent to FangBot servers.
What can an agent do on my machine?
Exactly as much as you allow. You choose when the agent must ask before running a command — Never, Allowlist, Ask, Auto-review or YOLO — and whether to sandbox it in Docker. File access is restricted to the agent's workspace.
How much does it cost?
The desktop app is free to download and use in local mode. You pay only for the model providers you connect — your own API keys, your own limits.
What happens when I quit the app?
Nothing stops. The agent runs as its own background service, so scheduled tasks and messaging channels keep working while the app is closed — and your data never leaves the machine.