As agentic AI becomes more prevalent in our workflows it's changing the touchpoints for how we create outcomes. I can brain dump into a note on my phone, pass it to an agent to clean up and catalog for me, then have another agent create a task in my Todoist with a first pass draft or sub-tasks to get me moving.
I've always developed at a laptop or desktop. Nothing beats typing at a keyboard when you're digging in and debugging lines of code. But as I've experimented more with agentic tools I've found myself wanting to send off a quick thought to an agent from wherever I am.
You can do that on Claude Code or Codex's web versions, but those don't get full access to your local machine. And they may not have the systems you've engineered to get the most out of your agentic coding tools.
It's those systems — the ones that define how models use the tools at their disposal — that really create the best work. And we want those on the go too.
This week we'll get into some of the options for taking your code with you so you can send off a prompt wherever inspiration may hit.
🔧 Tool
Claude Code Remote Control
Anthropic
The cleanest option if you're already in the Claude ecosystem. Run claude remote-control, scan a QR code, and your full local environment — MCP servers, tools, config — is on your phone. Nothing moves to the cloud and sessions reconnect if your laptop sleeps. One catch: the terminal process has to stay running. Close the lid and the session ends.
🎥 Watch
You Should Be Using Tailscale
Syntax.fm
The DIY version. Run OpenCode Web on a home machine, use Tailscale for a secure tunnel, and access from your phone. More moving parts, but you're not locked into any one tool. You could also run a Mac Mini (or other hardware) with local models. A nice developer setup to benefit from a beefy home machine while on the go. This one is on my to-do list to setup for myself.
🔧 Tool
Cmux
Not a "code on the go" tool, but it solves the adjacent problem — managing multiple agents without context-switching yourself into oblivion. Native macOS terminal built on Ghostty's engine. Works with Claude Code, Codex, OpenCode, Gemini CLI, and anything terminal-based. Free and open-source.
🗺️ Guide
Claw Camp
OpenClaw takes your entire agent team mobile. Interact through iMessage, WhatsApp, Telegram, Discord — all channels funnel to the same agent. Fire off a task on your commute, check back at your desk. Claw Camp is a free guide to setting up your first agent team from the creator of The AI Daily Brief. I'm working through it now and hoping to share my setup experience in coming weeks.
🧠 New Model
Introducing GPT-5.4
OpenAI
This breaks from the theme but it's the biggest news of the week. GPT-5.4 is the first general-purpose model with native computer use operating your machine via Playwright and keyboard/mouse commands. The remote tools above are about directing agents from a distance. GPT-5.4's computer use capabilities may handle some of that direction itself.
Thanks for reading,
Jason
