So I'm sitting at my desk, and on my right I've got my coding agent spinning up inside Orca. On my left, my phone is open and I'm about to type a prompt into it. I type something simple, like "hi", and hit submit. Immediately, the terminal on my desktop reacts. The change happens live, and I didn't touch my keyboard. That's the moment Orca clicked for me: I can now control my desktop AI agents from my mobile phone, and it's not just a gimmick. It's a fully functional interface that gives me access to everything I'd normally do in my terminal, plus a few things remote controls from Claude or Codex can't touch.
This open source repository lets you pair your mobile app with your desktop workspace, and it works across different coding agents. I've tested it with Claude Code, Open Code, Codex, and Harness agents, and the experience stays consistent. In this walkthrough I'll show you exactly how to install it, how to set it up, and most importantly, how it stacks up against the remote controls we've already been using.
What Orca actually is
Orca is a desktop application that acts as a hub for your AI coding agents. You spin up your agent inside it, manage your projects, split terminals, browse files, and run commands, just like you would in VS Code or a standalone terminal. The twist is that Orca also has a mobile app that connects to your desktop session. Once paired, your phone becomes a remote control that mirrors the agent's terminal, lets you type prompts, use hotkeys, switch between multiple computers, and even trigger custom skills with IntelliSense.
The repository is open source, so you can grab it directly from GitHub and install it on your local machine. It supports macOS, Windows, and Linux, and there's even a Homebrew option if that's your thing.
Installing Orca on your desktop
First, head over to the Orca repository and download the installer for your operating system. I downloaded the direct binary, opened it, and walked through a quick setup wizard.
The wizard asks you to pick your default coding agent. You can choose from Claude Code, Open Code, Codex, or any other agent you have installed. You can also expand the list later and switch between agents on the fly. At the bottom, there's a toggle for "YOLO mode" and "dangerously skip permissions". I left those off for now, but if you're comfortable running fully autonomous sessions, you can flip them on. Then you pick a theme (dark or light) and you're in.
Once you're inside, you add a project by browsing to a local folder. I opened a desktop project I had for Eric OS. Right away, Orca shows you a terminal window where you can type any coding agent command. You can also set up a default script that runs every time the workspace opens, like an npm install or a custom setup script. What's really special is that you can split the terminal view and run two or more agents side by side. I had Claude Code on one side and Open Code on the other, both working in parallel.
On the right sidebar, you have a file explorer that shows all the files and folders in your project directory. You can open any file, edit it, delete it, or run quick commands. There's also a source control panel that tracks changes, untracked files, and diffs, just like VS Code. So essentially, Orca gives you a full development environment that you can later control from your phone.
Pairing your mobile phone with Orca
The mobile app is where things get interesting. From the desktop app, you click on the mobile icon and choose your platform. I'm on Android, so I scanned the QR code or downloaded the APK directly. Once installed, you open the app and you're presented with two pairing options: Orca Relay and local network.
If your phone and computer are on the same Wi-Fi, you can use the local network option. But I wanted to be able to control my agents from anywhere, even outside my home network, so I chose Orca Relay. That required signing in with an Orca account, and after that, I could either scan a barcode from the desktop app or copy a pairing code and paste it into the mobile app. The pairing took seconds, and suddenly my phone was showing the same terminal session I had on my desktop.
Now, before I continue, a quick note: my name is Eric, I used to work as a senior software engineer at companies like Amazon and Microsoft, and I now run a school community where I help people master AI agents, automations, and building real SaaS products. We have courses, weekly live calls, and a bunch of resources. If that sounds interesting, you can jump in. But let's get back to Orca.
What you can actually do from the mobile app
Once paired, the mobile app mirrors your desktop terminal. I opened my Claude Code session and typed a prompt on my phone. On the left (my phone), I saw the agent start compiling, and if I tapped into the session, I could see exactly what was happening in the terminal, line by line.
What's really cool is that the mobile app supports IntelliSense for slash commands. If I type / on my phone, I see the exact same skill suggestions I have on my desktop. I can trigger simplified skills, superpower skills, or any custom skill I've downloaded locally. That means I can kick off complex workflows without touching my keyboard. I can also back out, open a different repository or workspace, and pretty much do anything I'd do on the desktop.
Hotkeys are fully supported too. I can do Shift+Tab to change permission modes, Escape, Enter, arrow keys, Delete, all from the mobile interface. That's something I haven't seen in other remote control solutions. And I can scroll down to see usage stats: my Codex subscription usage, my Claude usage, and even my GitHub tasks if I've connected a project.
Speaking of GitHub tasks, if you have a GitHub project board connected, you can see all your tasks right inside Orca's mobile app. I opened a project that had a Kanban board set up, and I could view every ticket. I've built an agentic building system where I offload tasks to the backlog, and then when I move a ticket to the "Ready" column, a specialist agent picks it up and starts building. The board has columns for Building, QA, and Reviewing, each with its own curated skills and dynamic workflows. The whole process runs autonomously: the builder finishes, the QA agent writes tests and verifies the work, then the reviewer checks everything and either sends it back for revision or merges the pull request. With Orca, I can open any ticket on my phone, see the description, and even write new tasks directly into the GitHub project, all while I'm away from my desk.
Managing multiple computers from one phone
One of the features I really like is that you can connect your mobile app to multiple computers. In my setup, I have an M2 machine as my host and an M5 machine also connected. On the mobile app, I can switch between them with a tap. I can open the M5 session, continue a conversation, and send a prompt like "what are the things we haven't scraped in the YT invest folder?" The agent on that remote machine starts compiling, and I can verify on the desktop that it's actually running. Meanwhile, I can go back to the M2 session and send a different prompt, like "remove the M4A audio file and commit everything." Both agents run in parallel, and I can bounce back and forth between them, resuming conversations and changing modes with hotkeys, all from my phone.
This multi-computer capability is huge if you're running different projects on different machines or if you want to keep a long-running agent session going on one computer while you start something new on another. You don't need to be sitting in front of any of them.
How Orca compares to Claude's and Codex's remote controls
Claude and Codex both offer remote control features, but they're limited in ways that Orca isn't. Right off the bat, with those remote controls, you can't see customized context window bars or how much context you've used so far. You can't use any custom skills. If you type a slash command, there's no IntelliSense, no skill suggestions. Hotkeys like Shift+Tab, Escape, or arrow keys are either absent or not customizable. You're also stuck with whatever permission mode the remote control gives you; you can't toggle bypass permissions or switch modes on the fly.
With Orca, I can open not just Claude but any AI agent I want. I can switch between Claude Code, Open Code, Codex, and others, all from the mobile app. I can also open a file browser and view all directories and files, something the built-in remote controls don't offer. The ability to use skills, hotkeys, and multiple agents makes Orca a much more complete mobile interface for coding agents.
Additional features: computer use, voice, and orchestration
If you dig into Orca's settings, there's even more you can do. There's a "computer use" option that, once installed, lets Orca's CLI control a browser on your behalf. You can also enable voice dictation if you don't want to type, and there's an orchestrations feature that lets you have multiple coding agents orchestrated by Orca itself. That means Orca can become an orchestrator that delegates tasks to different agents, each with its own specialty. This is still early, but the direction is clear: Orca isn't just a remote terminal; it's a platform for managing and coordinating AI agents from anywhere.
Points clés à retenir
- Orca is an open-source desktop app that pairs with a mobile app, letting you control AI coding agents from your phone.
- You can split terminals, browse files, edit code, and use source control, just like a full IDE, all accessible remotely.
- The mobile app mirrors the terminal, supports IntelliSense for slash commands, and lets you use all hotkeys (Shift+Tab, Escape, arrows, etc.).
- You can connect your phone to multiple computers and switch between them, running parallel agent sessions.
- Compared to Claude's or Codex's remote controls, Orca gives you full access to skills, permission modes, usage stats, and multiple agent types.
- You can view and manage GitHub tasks directly from the mobile app, integrating with autonomous agentic workflows.
- Additional features like computer use, voice dictation, and multi-agent orchestration are available in settings.
The first time I saw my phone trigger a real terminal command on my desktop without me touching anything, I knew this wasn't just a novelty. Orca gives you the kind of freedom that makes coding agents genuinely useful when you're away from your desk. You can check on a long-running task, fix a quick bug, or kick off a new build while you're grabbing coffee. And because it's open source and agent-agnostic, you're not locked into any single ecosystem. That's the kind of flexibility I think we're going to need as these agents become a bigger part of how we build software.