Let’s talk
ai era strategy

Claude Code Remote Control: Run a Project Thread on Your Computer

Claude Code Remote Control lets you drive a Claude Code session on your own computer from claude.ai or your phone. A Claude Code project uses it to run a thread on your machine when a task needs your local files or tools. Here is how to start it, and what six threads on our Mac taught us.

Adam Sandler7 min readUpdated
Claude Code Remote Control: Run a Project Thread on Your Computer

Claude Code Remote Control lets you drive a Claude Code session on your own computer from claude.ai/code or the Claude mobile app. The work runs on your machine, with your files, tools and signed-in accounts, while you type from anywhere. In a Claude Code project, it is how a thread runs on your computer instead of in the cloud. Our SEO project has done that six times on a Mac. Product facts below come from Anthropic's Remote Control docs, read November 18, 2026.

What is Claude Code Remote Control?

Remote Control connects claude.ai/code or the Claude mobile app to a Claude Code session running on your own machine. The code runs locally, your files never leave the computer, and your local MCP servers, tools and project settings stay available. The browser or phone is only the window you type into.

That makes it different from the two other ways people reach Claude Code away from the desk. A cloud session runs somewhere else entirely. Dispatch sends a task from your phone to the desktop app. Remote Control is a live session you can watch and steer from another screen.

How do you start Claude Code Remote Control?

Run claude remote-control (or claude rc) in your project folder to start a server that shows a URL and a QR code. Inside a running session, type /remote-control or /rc instead. Open the URL, scan the code with the Claude app, or pick the session with the green dot in the claude.ai/code session list.

  1. Sign in. Run /login and choose your claude.ai account.
  2. Start it in the right folder. cd into the project, then run claude remote-control. Accept the one-time "Enable Remote Control?" prompt and the workspace trust prompt.
  3. Connect. Open the URL it prints, press the spacebar for a QR code, or find the session in claude.ai/code.
  4. Make it the default if you like it. In /config, turn on "Enable Remote Control for all sessions" so every session can be reached without the command.

Server mode takes a few useful flags: --name labels the session, --spawn worktree gives each new session its own git worktree, and --capacity caps how many sessions one server runs (32 by default). The same command works from the Code tab of the desktop app and the VS Code extension.

Which plans include Remote Control?

Pro, Max, Ultra, Team and Enterprise, per Anthropic's docs on November 18, 2026. You sign in with your claude.ai account, because API keys are not supported. On Team and Enterprise an Owner turns it on in the Claude Code admin settings first. It does not work through Bedrock, Foundry or a custom API gateway.

How does a Claude Code project run a thread on your computer?

A Claude Code project is a coordinator that starts a thread for each task. Threads run in the cloud by default. When a task needs something only your computer has, the thread asks to run there, you allow it once for that folder, and Remote Control carries the session. You keep talking to the same thread in the project; the work just happens on your machine.

Our project runs this site's SEO. Its threads live in the cloud, but the SEO tool we use, OpenSEO, is self-hosted on Adam's Mac. Most days the cloud reaches it through a connector. When that connector fails, a thread runs on the Mac instead and reads the tool at localhost. Our November 9 to 15 weekly report came in that way: the cloud connector returned 502 errors, so a Mac thread pulled Search Console, GA4 and URL Inspection data locally, and the report shipped on time.

What is the difference between Remote Control and a cloud session?

Where the work runs. A cloud session runs on Anthropic's machines, needs no setup, and can run many tasks in parallel. Remote Control runs on your computer, so it can reach local files, local servers and tools you are signed in to. A Claude Code project uses cloud threads by default and Remote Control when a task needs your machine.

QuestionCloud threadThread on your computer
Where does the code run?Anthropic's machinesYour machine
Can it reach localhost or local files?NoYes
Does it need your computer on?NoYes, awake and running Claude Code
Does it read project memory?Yes, at the startInstructions yes, memory files no
Many tasks at once?YesLimited by your machine

Our rule: cloud by default, the computer only for what the cloud can't reach. That keeps the project running when the laptop is closed.

What did running threads on our Mac teach us?

  • A sleeping computer stalls the thread. The session reconnects after sleep, but nothing happens while the lid is shut. Our data pulls have waited on a sleeping Mac more than once. The fix was to keep the cloud as the default path and use the Mac as the fallback, not the plan.
  • Local permissions still apply. The Mac session runs under that machine's permission settings, not the cloud thread's. Once, its permission check blocked a save into OpenSEO that the cloud thread had planned. The report still landed in the project's shared files, and the save moved to the next run.
  • Memory does not travel. A thread on your computer starts with the project instructions but not the memory files. Anything a local thread must know goes in the instructions or the task itself.
  • Data first, then back to the cloud. Our Mac threads pull data and hand it back. Writing, building and pull requests stay in cloud threads, which run in parallel and never wait on a laptop.

Is Claude Code Remote Control secure?

Your machine makes outbound HTTPS requests only and never opens an inbound port. The session registers with the Anthropic API and polls for work, and all traffic passes through that API over TLS. Trusted Devices, in beta, can require each browser or phone to be enrolled before it can view or steer a session.

On Pro, Max and Ultra you turn on "Require trusted devices" yourself. On Team and Enterprise an Owner turns it on for the organization.

Why does local access matter for a second brain?

Most people's working knowledge lives on their own computer: notes, drafts, an Obsidian folder, exports nobody uploaded anywhere. A cloud session can't read any of it. Remote Control lets Claude work inside that material where it sits, which is what turns a pile of notes into a second brain you can query. For a team, the same idea scales into an internal knowledge base that lives somewhere every tool can read.

What are the limits?

  • The claude process has to keep running. Close the terminal or quit the app and the session goes offline until you resume it.
  • Outside server mode, one Claude Code process carries one remote session at a time.
  • In server mode, a network outage of about 10 minutes ends the server process. An interactive session keeps retrying until the network returns.
  • On a remote machine you SSH into, start claude inside tmux or screen so it survives your disconnect.

Where to start

Run claude remote-control in one project folder and send it a small task from your phone. If you run Claude Code projects for clients and want to package that into a service, the Claude Code Service OS covers the offer, pricing and outreach. Our Cowork vs Claude Code comparison helps if you are still choosing a tool.


Local threads write better when your brand is written down. In Brand Architect, Ophelia builds a knowledge base from your website or materials that any AI tool can read. Build your knowledge base →