At some point, one agent isn't enough.
It starts reasonably. You've got OpenClaw running, it's useful, it knows your projects. Then you want a second one — something isolated for experiments, or a work agent that doesn't know anything about your personal life. Fair enough. OpenClaw has profiles. You look it up, find the --profile flag, fire up a second terminal tab.
That works. Until it doesn't.
You've got two tabs open, maybe three. You can't remember which terminal is running which agent. You switch to the wrong one, start typing, realise you're talking to the dev agent instead of the personal one. You close the wrong window. You forget the port numbers. Two agents try to bind to the same port and one of them fails silently — and you only notice twenty minutes later when you can't figure out why nothing is working.
This is the actual experience of running multiple OpenClaw instances today. Not a documentation problem. Not a skill issue. Just friction that compounds every time you add an agent.

The frustrating part is that the underlying capability is already there. OpenClaw's profile system handles true isolation — separate memory, sessions, workspaces, API auth. Each profile is its own contained environment. The hard part isn't the isolation. It's the management layer on top. There isn't one.
That's the gap ManyClaw fills. A native macOS app that wraps OpenClaw's profile system in a proper interface — a sidebar of named, colour-coded agents, one click to start or stop, status visible at a glance. No terminal juggling. No port arithmetic. No guessing which tab is which.

Each agent stays fully isolated. Your dev agent doesn't know what your personal agent is working on. If one crashes, the others don't notice.
For the times you want the full experience, the interactive TUI is one click away — same window, same toolbar.

There's an in-app chat view too. Send a message to any running instance, read the reply inline. No browser, no context switching — fast enough that the plumbing disappears.

If you've ever sat with three terminal tabs open trying to keep track of which agent is which, you already understand the problem. The fix should have existed a long time ago.