Open beta
Your AI agents are working.
You don't have to babysit them.
Monitor and control Claude Code, Codex, Gemini CLI, Grok, and local Ollama models from your phone. Get push notifications when the agent needs you: approve or deny tool calls with one tap, send a new prompt mid-session, all end-to-end encrypted, from anywhere.
Download for macOSAvailable on iPhone and/or Android
Agents At Work supports
Not affiliated with Anthropic, OpenAI, Google, xAI, or Ollama.
Sounds exactly like what I wanted. I know some of the agents already support this, but not all to the same feature set and I bounce around between my agents a lot. Excited to see how this turns out!
Our phone is the sun of our personal device solar system, having the remote control there makes so much sense!
Run open-source AI models on your Mac.
Control them from your phone.
Agents At Work now works with Ollama, an open-source runtime for running AI models locally on your Mac. Pick a model, start a session, and send prompts from your phone exactly as you would with Claude, Codex, or any other agent. Any model works for chat, writing, and Q&A. Models in the qwen2.5 and qwen3 families also unlock Agentic mode, where the model reads and writes files, runs shell commands, and completes multi-step tasks on your Mac while you're away from it, the same workflow as Claude Code, but private and free.
The bigger your Mac, the more capable the model. An 8 GB Mac runs a 7 B model smoothly and it is genuinely useful for everyday writing, research, and planning tasks. A 16 GB or 32 GB Apple Silicon Mac can run 14 B or 32 B models that hold their own against cloud APIs for many real-world tasks. If you have a Mac Studio or Mac Pro with 64 GB or more of unified memory, you can run models that rival top cloud services, fully offline, as many times a day as you want.
Because everything runs on your hardware, your prompts never reach a third-party AI service. No API key. No per-query charges. No subscription beyond Ollama itself, which is free and open source.
Not just for developers
"I use it for writing posts and articles, planning our Europe trip, brainstorming, basically anything where I want an AI agent working on something while I'm away from my desk. Had an idea during a run last week, fired off a prompt from my phone, and came back to a full draft."
How it works
Want the full walkthrough? See the tutorial →
- Install the macOS companion app on the Mac where your agents run. Sign in, and your Mac is ready to bridge to your phone.
- Download the mobile app on iPhone (App Store) or Android (Google Play). No sign-in needed on the phone. Open it, go through the quick onboarding, and tap Scan QR Code. Point your camera at the QR shown in the Mac companion app, done. Your phone is linked, and the communication channel is end-to-end encrypted, in one step.
- Leave Claude, Codex, Gemini, Grok, or Ollama running. You'll get a notification the moment the agent needs a decision. Tap Yes or No right from the notification, or from your Apple Watch.
Control everything from your pocket
Four agents running in parallel. One tap to approve. Zero desk required.
Everything you need to step away
Push notifications with action buttons
When the agent asks to edit a file, run a command, or make a decision, you get an instant notification with Yes / No buttons. Answer directly from the lock screen, no need to unlock or open the app.
Live conversation feed
Every message, tool call, and assistant response streams to your phone in real time. A status bar shows "Wrote main.py", "Ran: npm test" between turns so you always know what's happening.
Five agents, one app
Claude Code, Codex, Gemini CLI, Grok, and Ollama are all fully supported. Each session shows a color-coded badge so you always know which agent is running and in which project.
Local models with Ollama
Install Ollama on your Mac and Agents At Work detects it automatically. Choose any model, start a session, and control it from your phone. Any model works for chat; the qwen2.5 and qwen3 families unlock Agentic mode so the model can act on your files and run commands, same as a cloud agent, but private and free. The bigger your Mac's memory, the more powerful the model you can run.
Auto-approve mode
Toggle "Allow all for this session" when you trust the agent to proceed without interruption. Resets automatically when you send a message or the session ends.
Send a message mid-session
Use the BTW field to nudge the running agent: corrections, new instructions, or extra context, without stopping its flow.
Multiple Macs and projects
Add as many computers as you like. Each shows detected agents, live sessions, and usage stats. Tap any project to open its conversation.
End-to-end encrypted
Conversation content is AES-256 encrypted on your Mac before it reaches Firebase. The encryption key is exchanged via a one-time QR scan, only your Mac and phone can read it. Firebase never sees plaintext.
Works on Apple Watch & Wear OS
Because the app uses standard iOS and Android notifications, permission prompts arrive on your watch automatically, no extra setup. Tap Yes or No on your wrist and the agent continues without you reaching for your phone.
Up and running in under a minute
Five quick screens. Tap the phone to walk through them. Want the full walkthrough? See the tutorial →
What you need
Mac
- macOS 13 Ventura or later
- Claude Code, Codex, Gemini CLI, or Grok, or Ollama for private local models
- Agents At Work companion app, download →
iPhone
- iOS 17 or later
- Available on the App Store
- Links to your Mac via QR scan, no account needed on phone
Android
- Android 8.0 (API 26) or later
- Available on Google Play
- Links to your Mac via QR scan, no account needed on phone
Have a proposal, or need to deploy this to a bigger team?
Happy to work with you directly on rollout, custom integrations, or whatever you have in mind. Let's figure it out together.
Get the macOS companion app
Free. Notarized by Apple. Direct download, no Mac App Store required.
Download AgentsAtWork.dmg- Open the downloaded
.dmgfile. - Drag Agents At Work to your Applications folder.
- On first launch, macOS may show a Gatekeeper dialog, click Open. The app is notarized by Apple.
- Sign in with Google or email/password and follow the in-app walkthrough.
Need to install an agent CLI (Claude Code, Codex, Gemini, Grok), set up Ollama for local models, or install tmux? Full setup walkthrough in the tutorial →