Your Agent's Memory, From Your Phone

On February 24, 2026, Anthropic released Claude Code Remote Control as a research preview. The idea is simple: run claude remote-control (or type /rc inside a session) in your terminal, and Claude Code generates a URL and QR code. Scan the code on your phone, and you are talking to your full Claude Code environment from a mobile browser.

The local process makes outbound HTTPS connections only. No inbound ports are opened on your machine. All traffic runs over TLS through the Anthropic API. Your filesystem, MCP servers, tools, and project configuration all stay available on the remote end, exactly as if you were typing in the terminal.

Remote Control is available today on Claude Pro ($20/mo) and Max ($100-200/mo) plans. One remote session at a time. Your terminal must stay open, but the session survives laptop sleep and reconnects automatically when your machine wakes up.

Here is the line that matters from Anthropic's announcement: “your filesystem, MCP servers, tools, and project configuration all stay available.” MCP servers are not stripped out for the mobile experience. They work.

If you have OMEGA configured as an MCP server in Claude Code, all 12+ tools are accessible through Remote Control. From your phone, you can:

This is the first time a persistent memory layer has been accessible from a mobile device without building a separate app. Remote Control turns Claude Code into the app, and OMEGA rides along for free.

Three steps. Nothing to install on your phone.

That is it. No port forwarding, no ngrok, no VPN. The outbound-only architecture means it works behind corporate firewalls, on hotel Wi-Fi, and on cellular connections.

Every MCP tool call requires manual approval. When you ask Claude to query your memory from your phone, you will see a confirmation prompt for the omega_query call and need to tap “Allow.” The same applies to omega_store, omega_checkpoint, and every other tool.

There is no auto-approve mode yet. Anthropic has been clear that this is intentional: Remote Control sessions carry real access to your filesystem and tools, and they want explicit consent for every action. There is a GitHub issue requesting read-only auto-approve for tools that do not modify state, which would make the mobile experience significantly smoother for queries. But for now, every tool call is a tap.

This is consistent with how Anthropic thinks about security. The session runs through their API over TLS, no inbound ports are opened, and every action requires your explicit approval. For a research preview, that is the right tradeoff. It will likely loosen over time as the feature matures.