Sam Altman called it last year: "building software will feel like playing StarCraft"

Someone actually built it. Tactical maps, minimap, idle worker alerts, for your AI agents.

The interface question just got real. Terminal tabs don't scale when you're commanding dozens of agents at once.

Website: getagentcraft.com

Got one agent in a chat window? You're fine.

Got 20 agents running async workflows and need to know who's stuck, who's idle, and what's on fire?

You're tab-switching into hell.

The bottleneck isn't model intelligence anymore. It's interface.

Ido Salomon built AgentCraft because terminals don't scale.

Website: getagentcraft.com

What it actually does right now:
- See every Claude Code agent on a single tactical map
- Summon new agents, assign missions, track progress live
- Idle worker alerts. RTS players know exactly what this means
- One command to install: npx @idosal/agentcraft

That's the v1.

It's rough around the edges. Documentation is basically nonexistent. Only a handful of people have actually used it.

But the shift is already here.

Full transparency. This is deeply early.

- Prototype stage
- Creator is dogfooding it with his own Claude Code setups
- No public docs, no GitHub, no API

One early user managed 9 agents with it and said "there's a lot to explore, but it feels right."

That's the signal.

We're heading toward a world where one human manages dozens of AI agents simultaneously.

The companies that figure out the interface first? They win.

→ Check the waitlist: getagentcraft.com

→ Think about it: What's your ideal agent UI? RTS? Kanban? Something we haven't seen yet?