Claude Code Remote Control: Access and Control Sessions from Any Device

The average Claude Code developer spends 20 hours per week inside the tool, and Claude Code now writes roughly 4% of all public GitHub commits. Those numbers reflect a workflow where the AI is a constant collaborator, not an occasional helper. But the default mode ties you to one terminal: close the lid, lose the session.

Claude Code Remote Control, released in research preview in early 2026, solves this by turning any browser or the Claude mobile app into a viewport for your local session. Your code, environment variables, MCP server connections, and project configuration all stay on your machine. The remote device is just a window into the running process, not a separate cloud environment.

This distinction matters. With VS Code Remote SSH or GitHub Codespaces, you move your development environment to a server. With Remote Control, the environment stays put. Your phone sends messages through the Anthropic API, and Claude Code on your machine executes them locally. No ports open inbound on your workstation. The connection uses outbound HTTPS only, the same transport as a normal Claude Code session.

Three practical scenarios make this valuable: reviewing a long-running refactor from your phone while away from your desk, kicking off a test suite on your workstation and monitoring progress from a tablet, and collaborating with a teammate by sharing a session URL while you both watch Claude work. Each scenario keeps your full local context intact, including any Fast.io MCP server connections, private repositories, and custom tooling.

Remote Control requires Claude Code v2.1.51 or later, a Claude Pro, Max, Team, or Enterprise subscription, and OAuth authentication (API-key-only setups don't work). On Team and Enterprise plans, an admin must enable the Remote Control toggle in the Claude Code admin settings before anyone on the team can use it.

Check your version first:

claude --version

If you need to update, Claude Code auto-updates by default. Otherwise run npm install -g @anthropic-ai/claude-code@latest.

There are three ways to start a Remote Control session, each suited to different situations.

Server mode dedicates the terminal to accepting remote connections:

claude remote-control --name "My Project"

This prints a session URL and lets you press spacebar to show a QR code. The process stays running and waiting for connections. Server mode supports multiple concurrent sessions (up to 32 by default) and can isolate each session in its own git worktree with --spawn worktree.

Interactive mode gives you a normal terminal session that is also accessible remotely:

claude --remote-control

You can type locally and from the remote device simultaneously. Messages stay synchronized across all connected surfaces.

On-the-fly upgrade converts an existing session without losing context. Type /remote-control or /rc inside any running Claude Code session. Your full conversation history carries over, and a session URL appears immediately.

To make Remote Control activate automatically for every session, run /config inside Claude Code and enable "Enable Remote Control for all sessions." Each claude process registers one remote session. If you run multiple terminal instances, each gets its own session.

Remote Control connects a browser to your local machine. Headless mode solves a different problem: running Claude Code programmatically on a CI/CD server, cloud VM, or any machine without a terminal attached.

Pass -p with your prompt to run non-interactively:

claude -p "Run the test suite and fix any failures" \
  --allowedTools "Bash,Read,Edit"

Add --bare to skip loading hooks, skills, plugins, MCP servers, and CLAUDE.md files. This reduces startup time and makes behavior consistent across machines:

claude --bare -p "Summarize this file" --allowedTools "Read"

Headless mode accepts piped input, so you can feed data through standard Unix pipelines:

cat build-error.txt | claude -p "explain the root cause" > diagnosis.txt

For structured output in automation scripts, use --output-format json with optional --json-schema to enforce a specific response shape. The JSON response includes session ID, cost data, and the result:

claude -p "Extract function names from auth.py" \
  --output-format json \
  --json-schema '{"type":"object","properties":{"functions":{"type":"array","items":{"type":"string"}}}}'

You can chain sessions using --continue or --resume with a session ID. This lets you build multi-step automation where each step picks up where the previous one left off:

session_id=$(claude -p "Start a review" --output-format json | jq -r '.session_id')
claude -p "Continue that review" --resume "$session_id"

The two approaches serve different needs. Remote Control keeps a human in the loop across devices. Headless mode removes the human entirely, letting automation scripts drive Claude Code through API keys and structured output. Use Remote Control when you want to monitor and approve work from your phone; use headless mode when a CI runner should handle everything without intervention.

Give your Claude Code sessions persistent, searchable file storage

A shared workspace with built-in AI indexing. Connect via MCP server, store agent output across sessions, and hand off to teammates. Starts with a 14-day free trial.

Start 14-Day Trial

Dev containers offer a third approach: running Claude Code inside a Docker container that provides a consistent environment for every team member. This matters for remote workflows because dev containers pair with cloud hosts like GitHub Codespaces, giving you a fully configured remote development environment without provisioning your own server.

Anthropic publishes an official Dev Container Feature for Claude Code. Adding it to your repository's dev container configuration installs Claude Code automatically when the container builds. Once inside, run claude in the terminal to authenticate and start working. See the official dev container documentation for the current setup steps and configuration options.

For teams, dev containers solve two problems that Remote Control and headless mode don't address. Environment consistency means every developer gets the same OS, dependencies, and tooling regardless of their local machine. Security isolation means commands Claude runs execute inside the container, not on the host. You can restrict network egress to only the domains Claude Code needs and configure unattended operation when the container is properly locked down.

Persistent authentication and organization-wide policy enforcement are both supported through container configuration. Teams can mount volumes to preserve login state across rebuilds and deploy managed settings that override individual developer preferences, ensuring consistent permissions and tool access across all sessions.

Claude Code offers several remote access patterns, and picking the wrong one wastes setup time or limits what you can do. Here is a decision framework based on what you're trying to accomplish.

Remote Control is the right choice when you're in the middle of local work and want to keep going from another device. Your filesystem, MCP servers, and environment stay available. The local process must keep running, so pair it with tmux or screen if your machine might sleep. Network outages longer than about 10 minutes cause a timeout, but shorter interruptions reconnect automatically.

Headless mode (claude -p) is right for automation, CI/CD pipelines, and scripted workflows. It works with API keys, doesn't need a subscribed user watching, and outputs structured JSON for downstream processing. Use it when no human needs to interact with the session.

Dev containers are right for teams that need reproducible environments, security boundaries, or cloud-hosted workstations. They pair well with GitHub Codespaces for fully remote development without maintaining infrastructure.

Dispatch lets you message a task from the Claude mobile app to your desktop machine. It requires the Claude Desktop app and mobile pairing, making it simpler than Remote Control for one-off task delegation.

Claude Code on the web runs entirely on Anthropic's cloud infrastructure. Choose it when you don't have a local clone of the repo, want to run multiple tasks in parallel, or need zero local setup.

For teams storing project files, build artifacts, or agent output that needs to persist across sessions and devices, Fast.io workspaces provide a shared layer. Files uploaded to a workspace are automatically indexed for semantic search through Intelligence Mode, so any team member or agent can query project documentation without manual organization. The Fast.io MCP server integrates directly with Claude Code, giving both local and remote sessions access to workspace files through 19 consolidated tools.

Local storage alternatives like S3 or Google Drive work for raw file storage. Fast.io adds the intelligence layer: upload a contract PDF, and any connected Claude Code session can answer questions about it with citations pulled from the file content. That capability persists whether you're accessing the session from your desk, your phone via Remote Control, or a headless CI runner.

"Remote Control requires a claude.ai subscription" means you're authenticated with an API key instead of a claude.ai account. Run claude auth login and choose the claude.ai option. If ANTHROPIC_API_KEY is set in your environment, unset it first.

"Remote Control requires a full-scope login token" appears when you've authenticated with a long-lived token from claude setup-token. These tokens only support inference, not Remote Control. Run claude auth login for a full-scope session token.

"Remote Control is disabled by your organization's policy" has several causes. On Team and Enterprise plans, an admin must enable the Remote Control toggle at claude.ai/admin-settings/claude-code. If the toggle is grayed out, your organization's data retention configuration is incompatible with Remote Control, and you'll need to contact Anthropic support.

Session disconnects when the laptop sleeps. Remote Control sessions reconnect automatically when your machine wakes up, but if the machine is awake with no network for more than roughly 10 minutes, the session times out and the process exits. Run claude remote-control again to start a fresh session. For long-running unattended work, wrap the command in a tmux session so it survives terminal closures.

Tool calls need approval from the remote device. Every tool call still requires your confirmation during Remote Control. Claude can't edit files or run commands without your approval. Front-load detailed instructions and edge cases in your initial prompt to minimize back-and-forth when you're away from a keyboard.

Session naming confusion on mobile. Sessions auto-name from the last message if you don't set a name explicitly. Run /rename your-session-name before activating /rc to give the session a recognizable name you can find quickly in the Claude app's session list.

Run claude doctor to diagnose issues that don't produce a clear error message. It checks authentication, subscription status, feature flags, and network connectivity in one pass.