OpenCode Gets a Mobile App: A Remote Control for the Agent on Your Computer
OpenCode now has an official mobile app, but the agent does not run on your phone. Sessions run on your computer, and the smartphone acts as a remote control. You can follow the agent's replies in real time and approve permission requests while you're away from your desk. To connect, run `opencode pair --remote` and scan a QR code that is valid for 5 minutes. For long-running agent tasks, it means you don't have to sit at the monitor.
AI-processed from TG: @neurogen_news (Neurogen); edited by Hamidun News
Coding agents have long been able to work for hours: refactoring a module, running tests, walking through a repository. The problem is that at the most inconvenient moment, the agent stops and waits for you to allow a command or answer a clarifying question. And at that moment you're in a café, on the subway, or simply away from your desk. OpenCode now has an answer to this problem: an official mobile app, OpenCode Mobile.
One important caveat up front: this is not an agent running on your phone. All sessions still run on your computer, and the app works as a remote control. The model, the tools, and the file access stay where they were. The phone only shows what is happening and relays your responses.
What the app can do
According to the author of the post, the feature set is as follows:
- view sessions and agent replies in real time and continue from where you left off;
- see which tools the agent has called and what it is working on right now;
- respond to permission requests and agent questions when you're away from your computer;
- attach images and dictate messages by voice.
The most practical item here is permissions. Imagine you've started the agent on a big migration and gone to lunch. Ten minutes later it asks for permission to run a command, and without your answer the work stands still. With the app, you tap a button on your phone and the task continues.
The project's website also promises voice calls: you call a session, speak out loud, the agent answers by voice, and at the end it tells you what it changed. But the app's Google Play description doesn't mention calls, only dictation. So for now, treat this feature as a promise rather than something you can use.
How to connect
The setup is simple:
- Install OpenCode v2 on your computer.
- Run
opencode pair --remote. It shows a QR code and a link, both valid for 5 minutes. - Scan the code in the app. Your recent sessions appear on the phone right away.
The --remote mode works through OpenTunnel, so you can connect from any network, not just your home Wi-Fi. If the phone and computer are on the same network, you can skip the tunnel: run opencode service set hostname 0.0.0.0 once, then simply run opencode pair.
The app is available on Google Play, and iPhone users can get a build through TestFlight. In other words, on iOS this is still a beta distribution rather than a regular App Store install. The project's source code is on GitHub in the anomalyco/opencode repository.
What to keep in mind
First, the computer running OpenCode must be on and reachable, otherwise the remote has nothing to control. If your laptop goes to sleep, the session won't respond.
Second, the original description says nothing about how the tunnel is secured or what permissions a paired device receives. Yet you are effectively letting a phone approve the agent's actions on a machine that holds your code. A QR code that lives for five minutes is a reasonable measure, but before using the app in a work environment, it makes sense to read the documentation and make sure you understand who can connect to a session and how.
Third, approving permissions from a phone is convenient, but it is easy to do without looking. If the agent is allowed to run commands in your repository, read each request on a small screen as carefully as you would at your desk.
Why it matters for developers
A mobile remote doesn't make the agent smarter, but it eliminates idle time. If you use OpenCode for long tasks, the app frees you from your workstation and lets you answer the agent's questions on time. It is worth trying for those who already work with OpenCode and have seen an agent wait hours for an answer. For everyone else, it is a good example of where coding agents are heading: execution stays on your machine, while control moves into your pocket.
- Nov 19, 2026Gemini API Can Now Load Tool Schemas on Demand Instead of in Every Prompt
- Nov 19, 2026Skyflow in AgentCore Gateway hides personal data from AI agents' tools
- Nov 19, 2026Chatbots cave after three objections: accuracy on climate answers drops 30–35%
- Nov 19, 2026Anthropic goes public and raises $81.3 billion, beating SpaceX's $75 billion record
Key Questions Answered
What the app can do?
According to the author of the post, the feature set is as follows: view sessions and agent replies in real time and continue from where you left off; see which tools the agent has called and what it is working on right now; respond to permission requests and agent questions when you're away from your computer; attach images and dictate messages by voice. The most practical item here is permissions. Imagine you've started the agent on a big migration and gone to lunch.
How to connect?
The setup is simple: Install OpenCode v2 on your computer. Run opencode pair --remote. It shows a QR code and a link, both valid for 5 minutes. Scan the code in the app. Your recent sessions appear on the phone right away. The --remote mode works through OpenTunnel, so you can connect from any network, not just your home Wi-Fi. If the phone and computer are on the same network, you can skip the tunnel: run opencode service set hostname 0.0.0.0 once, then simply run opencode pair.
What to keep in mind?
First, the computer running OpenCode must be on and reachable, otherwise the remote has nothing to control. If your laptop goes to sleep, the session won't respond. Second, the original description says nothing about how the tunnel is secured or what permissions a paired device receives. Yet you are effectively letting a phone approve the agent's actions on a machine that holds your code.
Test Frontier LLMs & Autonomous Agents in One Workspace
Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.