REM:Code is a mobile-first workspace for developers working with coding agents. Your agents run where your code and toolchain already are; your phone becomes the place to direct them, inspect the work, run commands, preview results, and install builds.
Ask for a build. The agent runs Xcode or Gradle on your machine and replies with an Install card. One tap, no TestFlight, no cable.
HTML in your project opens as a rendered page. Screenshots the agent takes appear right in the conversation. Reports and docs too.
Browse files, read per-file diffs, run a command in the terminal, commit — the parts of your IDE that fit on a phone.
Send tasks by text or voice, approve what needs approving, and get a push notification when it’s your turn.
Use Claude Code or OpenAI Codex per session. Keep the same mobile workflow instead of tying your projects to one model vendor.
Docker, Kubernetes, Xcode, your signing, your databases — everything stays where it’s already set up. Nothing runs in our cloud.
One agent on your machine. One app on your phone. A relay in between that only forwards.
One command on your laptop or server. The agent opens an outbound tunnel — no ports, no VPN, no cloud VM.
Claude Code or OpenAI Codex runs on that machine: your Docker, your Xcode, your certificates, your databases. Nothing is copied to our cloud.
Direct the agent, read the diff, open the page, install the build. Approve when it asks, get a push when it’s your turn.
Describe the feature, approve the edits, install the build, try it, send the next round. Swift, Kotlin, Flutter — whatever xcodebuild or Gradle produces on your Mac.
Mobile workflow →Diffs, files, rendered pages and screenshots — see what the agent actually did, then commit. For backends, the terminal and your reports are right there.
Reviewing from the phone →A laptop, a home server, a Windows box over RDP — one list of sessions, one app. Each session picks its own project, model and permission mode.
Sessions across machines →Because building from your phone needs more than pixels or messages — and shouldn’t need someone else’s cloud.
Your monitor shrunk to six inches, or text in and text out. Nothing designed for a thumb: no approvals, no diffs, no rendered pages, no Install button.
Their machines, their framework, their agent, credits per iteration. Your code lives in their cloud and the “app” runs inside their shell.
Your machine, your stack, your agent. A phone-sized view of everything it produces — code, pages, screenshots — and a real signed build in your hand.
Your machine opens an outbound WebSocket tunnel, so you do not expose inbound ports or configure a VPN. Connections to the relay are protected with TLS, and the relay is designed to forward interactive session traffic without persisting your code, prompts, or responses.
Your repositories, local session history, tools, credentials, and build environment remain on your machine.
How the connection works →A mobile-first development workspace for working with AI coding agents on your own machine. The phone app lets you direct them, inspect files and diffs, use the terminal, see rendered pages and screenshots, commit changes, and install the app you are building.
No. REM:Code is a professional tool for developers who already have repositories, toolchains, and coding agents. It makes the agentic development loop usable from a phone; it does not hide the code or replace engineering judgment.
Yes. Tell the agent to build it and put it on your phone. It exports an .ipa or .apk with your own signing, uploads it encrypted, and replies with an Install card. iOS installs over the air, Android hands the APK to the installer. Your Apple developer setup stays yours — REM:Code moves the build, it does not sign it.
HTML files in your project open as rendered pages inside the app. Screenshots the agent takes appear in the conversation. Markdown, images and code open in the file browser, and per-file diffs show every change with syntax highlighting.
Claude Code and OpenAI Codex, chosen per session. Whichever CLIs are installed on the machine show up automatically, and streaming, approvals, history and interrupts behave the same for both.
No. The machine keeps its Docker, Kubernetes, databases and everything else already set up, and the agent uses all of it. You direct, review diffs, run the terminal, read reports from your phone. Installing the build on the phone is just the most visible case.
No. Every build, test and edit runs on your machine. The gateway relays traffic, handles sign-in and push, and stores shared builds encrypted with a key that lives only in the link.
A remote desktop gives you pixels, a bot gives you text. Neither has tool approvals, per-file diffs, rendered pages or an Install button. REM:Code is built for the phone: what the agent produces arrives in a form you can read and act on with one thumb.
Connections from your phone and your machine to the relay are protected with TLS. The relay forwards interactive session traffic and is designed not to persist your code, prompts, or responses. See the Security page for the current architecture and threat model.
REM:Code is free during early access, including remote coding sessions, project review, voice input and notifications. A $4.99/month subscription is planned for a future release. No subscription or payment is required today. Your coding tool and AI model access are separate.
REM is the phase of sleep where the mind stays active. Your laptop can be closed, but your agents keep working — and the result is waiting on your phone.
Close the lid. Let them dream.