OpenClaw: 100 Agents, $1.3M in Tokens, 30 Days
The fastest-growing open-source project in GitHub history was built by ~100 AI agents running in parallel — at a $1.3M monthly token bill. The viral story hides the real lesson: orchestration and cost discipline, not raw model speed.
NeuroX AI · June 6, 2026
OpenClaw crossed 302,000 GitHub stars by April 2026 — the fastest any open-source project has ever reached that mark, overtaking React, Vue, and TensorFlow in a fraction of the time. The part the headlines skip: it was built by roughly 100 AI agents running in parallel.
That fleet wasn't free. Peter Steinberger's agents burned $1.3 million in tokens over 30 days — 603 billion tokens across 7.6 million requests. One person, orchestrating a swarm, spending like a mid-size engineering org.
Here's the number that should change how you think about it: the same output dropped to roughly $300K/month once Fast Mode pricing was switched off for work that didn't need it. Same agents, same code, a 4x swing — decided entirely by orchestration choices, not model capability.
That's the whole lesson. Spinning up 100 agents is a config flag now. Knowing which tasks deserve premium inference, which run on cheaper passes, where the review gates sit, and what the per-run cost actually is — that's the engineering. The teams burning money are the ones who scaled the fleet before they built the instrumentation.
Parallel agents are commoditized. The discipline around them is the moat.