Claude-Relay vs. Claude Code Remote Control

[seidnerj]

Remote Control does not (yet?) work without a subscription, i.e. using api keys, but I assume that is coming soon. Does this make this project obsolete?

https://x.com/noahzweben/status/2026371260805271615
https://code.claude.com/docs/en/remote-control

[chadbyte]

Great question, and honestly one I've been thinking about since Remote Control dropped.

The short answer: this frees us to go deeper.

Right now, claude-relay does more than Remote Control. Multi-session, API key support, Team/Enterprise compatibility, data that stays on your network. Some of these may always be differentiators. Some Anthropic will absorb. That's fine.

What I'm actually excited about is that Remote Control validates the problem we've been solving. Now that Anthropic owns the basic use case, we're free to build for power users. The casual "check on Claude from my phone" user is covered. Good.

That leaves us to focus on the analogy I keep coming back to:

• bash = Claude Code. The core.
• ssh = Remote Control. Connect from outside; disconnect and it's gone.
• tmux = claude-relay. Sessions persist. Detach, reattach, multiplex.

The goal isn't to compete on basics. It's to build the workspace layer that power users actually need.

Will Anthropic build this natively someday? Maybe. But open-source workspace tooling tends to stay open-source. tmux didn't die when terminals got tabs.

[chadbyte]

This question aged well. I pinned it because it's the most important question this project has faced, and I think the answer says a lot about where Clay is headed.

When Remote Control dropped, I seriously considered shutting this down. The original answer above was me trying to convince myself there was still a reason to keep going.

Let me be honest about something. The remote access feature is what got this project traction. I leaned into it, promoted it that way, and it worked. People came for "Claude Code on your phone." That's real and I'm not going to pretend otherwise.

But it's not why I built this. I had been putting my own context on top of Claude Code, making it into something more like a colleague than a tool. Claude Code itself becoming my AI, with my context layered on. I built claude-relay because I wanted to access that AI from my phone. The remote access was just a means to stay connected to it. The core was always the AI colleague.

Now that Remote Control covers the basic remote access use case, I don't need to lean on that anymore. It freed me to go where I actually wanted to go: AI as a teammate, not a tool.

So I rebranded to Clay and went all in. Every piece of it works on its own. The Slack-style workspace UI, multi-user with RBAC and invite codes. A visual scheduler for overnight agent runs. Ralph Loop, where a coder and judge agent iterate autonomously on your code. These are all real, useful features whether or not you care about the "AI colleague" angle.

I wanted to build a proper collaborative space for people first, then naturally layer in AI teammates on top of that. That's why Mates came after the workspace UI and multi-user. When you add an AI teammate to a space that already works like a team workspace, it lands as a colleague, not a feature.

That's the part that's starting to take shape now. You create a Mate through a conversation. Interview them, give them a name, an avatar, a personality. They remember context across sessions. You can DM them. They show up in your sidebar like any other colleague.

Remote Control and Clay solve different problems. I'm not worried about overlap anymore.