How Claude Code's New Auto-Memory and Remote Control Features Stack Up Against OpenClaw

What Changed — A Rapid Response to Competition

A month after a detailed comparison with the open-source agent OpenClaw, Claude Code has shipped significant updates. The most notable is the promotion of auto-memory to a stable feature, directly addressing what was identified as a major weakness. Furthermore, Claude Code now offers remote session control via Telegram and Discord channels, allowing you to steer a coding session from your phone.

What It Means For You — Concrete Impacts on Daily Use

These updates change how you can delegate work. Auto-memory means Claude Code should now retain context about your project and preferences across sessions, reducing repetitive explanations. The remote control feature enables a new workflow: you can start a long-running task (like a refactor or test suite) on your desktop and monitor or redirect it while away.

However, the source reveals critical limitations you must know. Remote control is not retroactive. You must explicitly enable it (claude code --enable-remote) at the start of a session. If you forget, you cannot connect later. This is a stark contrast to OpenClaw's persistent, always-available agent model.

Regarding OpenClaw, the landscape has also shifted. Anthropic clarified that using a Claude subscription with OpenClaw violates its Terms of Service. The recommended path is now to use the Agent Communication Protocol (ACP), where you can spawn Claude for specific processes. For developers who valued OpenClaw's autonomy with Claude's intelligence, this creates friction and can change the agent's "feel."

Try It Now — Commands and Configuration

To leverage the new features immediately:

1. Enable Auto-Memory: It should be on by default. Verify and configure its scope in your ~/.config/claude-code/config.yaml:

   memory:
     enabled: true
     # Options: session, project, global
     persistence_scope: project
   

2. Start a Remotely Controllable Session: Use the --enable-remote flag and specify a channel. You'll get a connection link.

   claude code --enable-remote --channel telegram
   

   Save the provided link to your phone. Remember, this only works for tasks started after this command.

3. For Heavy OpenClaw Users: If you're exploring ACP with Claude, structure your prompts to delegate discrete tasks to spawned Claude instances, rather than expecting it to be the primary, always-on brain.

The Bottom Line on Practicality

The source author's conclusion is telling: "The OpenClaw I am using today feels much less practical than the one I was using a month ago." For Claude Code users, the takeaway isn't about which tool "wins." It's that Claude Code is rapidly iterating by adopting proven concepts, but often with a more guarded, security-conscious implementation that trades some autonomy for polish and safety. Your choice hinges on whether you prioritize Anthropic's integrated, streamlined experience or are willing to manage the complexity and cost of an open-source stack for deeper autonomy.