Why We Built TetherClaw

Running an AI agent business means being chained to your desk. Here's the insight that changed how we thought about it — and what we built.

Every operator I know has the same problem. You're running AI agents that work around the clock — research pipelines, customer ops, development workflows — and the moment you step away from your desk, you're flying blind.

Your agents are running. Tasks are completing. And when something needs your input, the session just... hangs. Or worse, times out.

I built TetherClaw because I got tired of losing context every time I walked away from my machine.

The problem isn't the agents

AI agent platforms have gotten genuinely good. OpenClaw, Hermes Agent, CrewAI, LangChain — these tools can handle real business workflows now. The issue is that all of them assume you're sitting at a terminal.

The session is tied to the device. The device is tied to the desk. When you pick up your phone and try to continue — you're starting over. Re-explaining where things stood. Rebuilding context the machine should already have.

You end up being the sync. Manually carrying what the technology should carry.

You can duct-tape solutions together. Tailscale if you know what you're doing. ngrok for the desperate. SSH over a hotspot if you're truly committed. But all of these feel like what they are: workarounds for a problem that shouldn't exist.

The insight: session belongs on the relay, not the device

The fundamental shift in how I thought about this: the session shouldn't live on your laptop. It should live on a relay that your laptop, phone, and browser all connect to.

This isn't a new idea in networking. It's how SSH multiplexers work, how VDI works. But nobody had built it cleanly for the AI agent layer — with persistent memory, push notifications, and a mobile interface that actually works.

When the relay owns the session, everything changes. You disconnect from your laptop — the session keeps running. You open the app on your phone — the session is right there, mid-thought. You switch from WiFi to cellular in the elevator — nothing drops. Your agents remember the full context of what you've been working on, regardless of which device you're on.

That's what we built.

What TetherClaw is

TetherClaw is a universal mobile access layer for AI agent platforms. It sits between your agent runtime and your devices — any device, any network — and keeps the session alive and accessible no matter where you are.

You install the relay on whatever machine runs your agents. TetherClaw handles the secure tunnel, the session continuity, the push notifications. You get a native iOS app that connects to your agents like they're in the same room.

No Tailscale. No port forwarding. No configuring firewall rules at 11pm.

Who it's for

I want to be honest about who this is and isn't for.

If you're running agents experimentally — occasionally, locally, for personal projects — you probably don't need this yet. Your current setup is fine.

TetherClaw is for operators running real workloads on AI agents. People who have agents doing customer research, code review, content ops, support triage — tasks that run throughout the day and need a human in the loop at key decision points.

For those people, being unreachable isn't a minor inconvenience. It's a broken workflow.

Where we're going

The vision is simple: one app, every platform, anywhere.

We launched with support for OpenClaw and Hermes Agent because those are the platforms our first users run. But the relay protocol is platform-agnostic. We're adding more platforms on an ongoing basis based on where operators are actually running workloads.

The goal isn't to replace your agent platform. It's to make every agent platform actually portable.

If you're running agents and you recognize this problem, we'd love to have you try it.

— The TetherClaw team