Free · No account · iOS and Android
Your coding agent runs on your computer. You run it from your phone.
Leave the desk and keep your coding work moving. Coordinate agents, see tools, files and diffs as they work, answer questions, approve plans and put their results into your project — all from your phone. Your computer does the work; you stay in command.
Now available on the App Store and Google Play.
- No sign-up, no profile
- Pinned to your own certificate
- Your data never leaves your computer
How it works
Install, pair, send a task
One install command sets up the background service and gives you a pairing QR code. Scan it, then run your coding work from your phone while your computer handles the execution.
-
Install Agent Commander
Run one command on your computer to install Agent Commander. Before sending your first task, install and sign in to any one of these: Claude Code CLI, Codex CLI, or Antigravity. The installer picks the right package, checks it and starts setup for you.
-
Scan the pairing code
Scan the QR code to connect your phone to your computer and pin its certificate. Your app now knows which computer to trust.
-
Keep work moving from your phone
Send a request with a photo or file, inspect the work and answer the agent. Enable notifications to hear when results are ready or a decision needs you. The agent keeps working when you close the app.
How Agent Commander compares to other remote workflows
| Capability | Agent Commander | SSH / Termux | Remote Desktop (VNC) | Cloud IDE |
|---|---|---|---|---|
| Where code & agents run | Your own computer (agentcmd + SQLite) |
Your own computer | Your own computer | Third-party cloud VM |
| Mobile interface | Native iOS & Android UI for chat, diffs, and approval cards | Tiny terminal grid and mobile keyboard | Desktop screen streaming, high bandwidth | Mobile web browser |
| Tool approvals, plans & diffs | One-tap interactive cards, unified diff viewer & file tree | Raw terminal text scrolling | Pinch-zooming desktop windows | Web editor dependent |
| Multi-agent coordination | Claude Code, Codex, Antigravity + isolated Git worktree delegation | Separate manual shell sessions | Manual window switching | Usually locked to one vendor |
| Terminal CLI session handoff | Watch live Claude CLI, read Codex CLI history & branch to mobile | Only if pre-wrapped in tmux / screen |
Mirrors open desktop window | Disconnected from local terminal |
| Privacy & pricing | Free (0 USD), no account, SHA-256 certificate-pinned TLS | Manual SSH key management | Requires remote desktop account | Cloud compute subscription; code stored in cloud |
See the full technical fact sheet on the Press & Product Fact Sheet page.
Features
Your coding work, under your control wherever you are
Built around one idea: leaving the desk should not mean leaving the agent unattended.
-
See the work as it happens
Follow tool calls, files read or edited and diffs as the agent works. Open project file links straight from its reply. Type @ to point to a file, and use / in Claude conversations to run a skill.
-
Make decisions without returning to your desk
Answer questions, approve tool runs and review plans from your phone. Choose an answer or write your own. In Claude conversations, send another message during a running turn or queue the next task.
-
Bring your coding agents together
Coordinate Claude Code, Codex and Antigravity in one app. Choose Claude Code, Codex, or Antigravity for a conversation, then pick its model and effort. You only need one installed agent to get started.
-
Divide the work and see every result
Set project rules or pick an agent for a turn. Use separate Git branches for parallel work. Follow each managed sub-task’s progress, open its conversation, answer questions and read its report. The main agent reviews results, applies accepted work and sends feedback for another round.
-
Continue work from your terminal
Follow Claude CLI sessions live and read Codex CLI history from your phone. Branch a closed session into a new app conversation to take the work further, with its history at hand and the original session preserved.
-
Control, project by project
For Claude projects, choose ask first, accept edits, plan first or full access. Give each project its own system prompt, allow and deny tool rules, MCP servers, agents and skills — so the agent works by your rules.
-
Put the output to work
Files the agent produces land in one place. Preview images, SVG, HTML, Markdown, JSON and text right in the app, then save or share them. Export selected files into your project with conflict checks before replacing anything.
-
Browse and edit the project
Open the file tree, read code with syntax highlighting, edit and save, create, rename, move or delete. Git badges show what changed, and every file's history is one tap away.
-
Review the work
Commit history as a branch graph, the diff of every commit and every file, pull, push and publish a branch — all from the phone.
-
Keep a task list
Note what needs doing in each project, attach files, add comments, and link a task to the conversation that handles it. Tasks live on your server, so they survive a reinstall.
-
Know where your tokens and budget go
See today’s token total, token use and recorded cost per turn and per agent, plus project totals. Check plan limits and set reset reminders. Claude conversations show cache reads and writes, warmth and expiry, and estimated tokens to write again on the next turn.
-
Find the work that matters
Search conversations, filter by project or unread activity, and pin the ones you need. The notification log keeps results and items needing your attention together.
-
See inside the agent’s work
See Claude’s native sub-tasks in activity cards: status, duration, tool activity and token use. Know what each one is doing and where the effort is going.
-
Try another approach with your history intact
In Claude conversations, restore files to before a turn, branch the conversation to explore another approach, or compact context to keep going. The context meter shows when it is time to make room.
Screens
What it actually looks like
Screenshots from the app itself — English and Vietnamese, iPhone and iPad.
Install
One binary, one SQLite file, your own computer
Keep your coding infrastructure in your hands. Agent Commander runs in the background on the computer you already code on, stores your conversations there and gives your phone control of the agents working in your project.
# macOS · Linux · WSL — install and start setup
$ curl -fsSL https://github.com/phidinhtho/agentcmd/releases/latest/download/install.sh | bash
# need the pairing QR code again?
$ agentcmd auth show# Windows 11 — install and start setup in PowerShell
PS> irm https://github.com/phidinhtho/agentcmd/releases/latest/download/install.ps1 | iex
# need the pairing QR code again?
PS> agentcmd auth showThe installer detects your platform, verifies the release checksum, puts agentcmd where your account can run it, and hands over to setup.
What setup does for you
- Creates the data directory and a migrated SQLite database
- Generates the TLS key and certificate that the app will pin
- Writes an environment file readable only by you
- Installs and starts a background service (launchd, systemd or Task Scheduler)
- Prints a pairing QR code, valid for ten minutes
- One binaryA lightweight Go binary and a SQLite file. No database service, Node or Redis to maintain alongside it.
- Where it runsmacOS on Apple Silicon and Intel, Linux on amd64 and arm64, Windows 11 on x64, and WSL.
- Works with your agentClaude Code CLI, Codex CLI, or Antigravity. Install and sign in to any one of the three on the same computer.
- Stays healthydoctor checks the setup, status reports it, and update replaces the binary and restarts the service in place.
- Several apps, one computerPair several app installations with one computer, see them all and revoke each one.
- A desktop companionA menu bar app on macOS, a system tray app on Windows: start, stop and watch the server without opening a terminal.
Privacy
Your conversations, on infrastructure you control
Your computer holds your conversation history and project files. You choose the agents, run the software and control access. The publisher does not keep a copy of your conversations.
- No account, everNo sign-up, no profile, no third-party login. Pairing with your server is the only identity there is.
- Keep ownership of your workYour project folders and conversation database stay under your control. Access and deletion are in your hands, on your own computer.
- Pinned connectionsThe app records your server certificate the moment you pair, and refuses to talk to anything that does not match it afterwards.
- Notifications without sharing the conversationThe push relay delivers alerts without carrying conversation content or storing your work.
- No ads, no trackingNo in-app purchases, no advertising and no usage analytics. Anonymous crash reports only.
- Remove one deviceRemoving a paired device deletes only its associated local app data and downloaded files. Other devices and the data on your computer remain.
FAQ
The questions people actually ask
If yours is not here, the support page goes deeper.
Can I pair more than one computer?
Yes. Pair several computers, switch between them and give each a name. Removing one device deletes only the app’s local data associated with that device, including downloaded files; other paired devices and data on the computer remain.
Do I need an account?
No. There is no sign-up and no profile. The app pairs once with a server you run yourself, and that pairing is the only identity involved.
Which coding agents does it work with?
Agent Commander officially supports Claude Code, Codex, and Antigravity. To start a conversation, you only need to install and sign in to one of the three on your computer.
Where does my code live?
Your project folders and conversation database live on your computer, under your control. The publisher does not keep a conversation copy. See the Privacy Policy for how your chosen AI provider processes the content you send.
Can I reach my computer from another network?
Yes, through Tailscale or a public address. The pairing code carries every address your server has, local network included, and when one stops answering the app moves to another on its own — no new scan.
Can I pick up work I started in the terminal?
Yes. Follow Claude CLI sessions live and read Codex CLI history. From a closed session, branch into a new app conversation and keep working from your phone, with the terminal history available to read and the original session preserved.
What does it cost?
The app is free: no in-app purchases, no subscription and no ads. Whatever your coding agent costs to run is billed by your existing plan with its provider, not by us.
Which platforms are supported?
Now available on the App Store and Google Play. The app runs on iPhone, iPad and Android. On iOS, iOS 15 or later is required. The server runs on macOS (Apple Silicon and Intel), Linux (amd64 and arm64), Windows 11 (x64) and WSL.
What if a turn takes twenty minutes?
Close the app and carry on with your day: the agent keeps working on your computer. Enable notifications to know when results are ready or the agent needs a decision. The support guide walks you through turning them on.
Make your phone the command center for your coding work.
Install with one command, scan the QR code and send your first task. Leave the desk with your agents, decisions and results still in hand.
Now available on the App Store and Google Play.
Agent Commander is an independent, third-party client. It is not affiliated with, endorsed by, or sponsored by Anthropic. Claude and Claude Code are trademarks of Anthropic.