Remote Control and Claude Code on the web both live at claude.ai/code, which is exactly why they get confused. One difference explains everything else: where the code executes. Remote Control runs on your machine. Claude Code on the web runs on Anthropic-managed infrastructure, working from a GitHub clone.
Fast rule. Work already started on your machine and needs your real files: Remote Control. Work not started, or on a repository you have not cloned: the web.
Side by side
| Remote Control | Claude Code on the web | |
|---|---|---|
| Where it runs | Your machine (CLI or VS Code) | Anthropic-managed cloud VM |
| Local filesystem | Full access | None. Cloned from GitHub |
| Your MCP servers and tools | Stay available | Cloud tooling only |
| Project configuration | Your local config applies | Configured through cloud environments |
| Start it with | claude remote-control or /rc | claude --cloud "task" or the browser |
| Machine must stay on | Yes, it is a local process | No |
| Network outage | Times out after roughly 10 minutes | Unaffected |
| Repository you have not cloned | Needs a local clone | Works without one |
| Parallel tasks | Server mode, up to 32 sessions | Each task is its own session |
What each one genuinely cannot do
The comparison table lists differences. These are the hard stops, and they usually settle the choice faster.
Remote Control cannot
- Survive your machine being off, asleep past the timeout, or the terminal closing.
- Run without a claude.ai login. API keys and long-lived tokens are rejected.
- Work through a proxy or gateway. A custom
ANTHROPIC_BASE_URLdisables it, as do Bedrock, Google Cloud Agent Platform and Microsoft Foundry.
Claude Code on the web cannot
- See a file you have not pushed. The VM clones from GitHub, not from your disk.
- Reach a local MCP server, a local database, or anything else on your network.
- Push results anywhere but GitHub. GitLab and Bitbucket repositories can be uploaded as a bundle, but the session cannot push back.
- Run at all under organisation IP allowlisting, because sessions call the API from Anthropic infrastructure rather than your network.
One constraint blocks both: organisations with Zero Data Retention cannot use cloud sessions, and cannot enable Remote Control either. If that is your situation, neither of these is the answer.
Moving work between them
They are not mutually exclusive, and the handoff is the part most people miss.
Terminal to cloud
claude --cloud "Fix the authentication bug in src/auth/login.ts"The catch on a first run: the VM clones your current branch from GitHub. Local commits you have not pushed do not exist as far as that session is concerned. Push first.
If the repository has no GitHub remote, Claude Code bundles it and uploads it directly. That fallback has limits worth knowing: the bundle must be under 100 MB, untracked files are excluded so git add anything the session needs to see, and a bundled session cannot push back to a remote.
Cloud back to terminal
Pull a cloud session down with claude --teleport for a picker, /teleport or /tp inside a session, or /tasks then t. Teleport checks four things first: a clean git state, the correct repository rather than a fork, the branch pushed to the remote, and the same claude.ai account.
The overlap that confuses people. --teleport runs on the same session infrastructure Remote Control uses, so teleport failures surface with Remote Control wording. Remote Control session expired or Remote Control may not be available for this organization while teleporting is a cloud-session access problem, not a Remote Control fault. Detail in Claude Code on the web.
And back to your phone
After teleporting, the terminal gets its own copy and new work stops appearing in the cloud session. To keep steering from your phone, start Remote Control in the local session with /remote-control. That is the clean way to use both: cloud for the long unattended run, teleport to bring it home, Remote Control to keep watching.
A worked sequence
The pattern that uses each tool for what it is good at:
- Start locally in plan mode so Claude explores and proposes an approach without editing anything.
- Agree the plan, then commit and push it, because the VM clones from GitHub.
- Hand it off:
claude --cloud "Execute the migration plan in docs/migration-plan.md". - Close your laptop. The cloud session keeps going.
- Teleport it back when you want to finish locally, then
/remote-controlif you want to keep an eye on it from your phone.
Cost and limits
Neither charges separately for compute, but cloud sessions share rate limits with all your other Claude and Claude Code usage. Running several --cloud tasks in parallel consumes them proportionately, which is the practical ceiling on how many you run at once.
Cloud sessions also expire after a period of inactivity and their VM is reclaimed. Reopening from claude.ai provisions a fresh one with your conversation history restored, so nothing is lost, but a long-idle session is not instantly ready.
Isolation, if that is what decides it
Each cloud session runs in its own isolated Anthropic-managed VM. Network access is limited by default and can be disabled entirely, though Claude Code still reaches the Anthropic API even then, which means data can leave the VM by that route. Git credentials and signing keys are never inside the sandbox; authentication goes through a proxy using scoped credentials.
Remote Control inverts this. Nothing about your code leaves your machine, but the session transcript is stored on Anthropic servers while connected. Which of those you prefer depends on whether you are protecting the code or the conversation. See the security model.
The wider set of options
These two are not the only ways to reach Claude Code away from the terminal.
| Option | Trigger | Claude runs on | Best for |
|---|---|---|---|
| Dispatch | Message a task from the Claude app | Your machine, via Desktop | Delegating with minimal setup. Pro and Max only |
| Remote Control | Drive a running session from claude.ai or the app | Your machine | Steering work already underway |
| Channels | Events pushed from Telegram, Discord, CI | Your machine | Reacting to external events |
| Slack | Mention @Claude in a channel | Anthropic cloud | Starting work from team conversation |
| Scheduled tasks | A schedule you set | CLI, Desktop, or cloud | Recurring automation |
Can you run both at once?
Yes, on different work, and there is no longer a conflict between them. The Ultraplan preview, which used to fight Remote Control for the claude.ai/code interface, was removed in Claude Code v2.1.224. Plan mode and Claude Code on the web replace it.
More in remote workflows, and on always-on machines if the real problem is that your laptop keeps closing.