runstate - Agent Swarm Coordination
The coordination backbone for the future of agent swarms
4 followers
The coordination backbone for the future of agent swarms
4 followers
runstate gives agent swarms one shared place for work, quotas, budgets, and results across processes and machines. Agents can claim tasks without duplicating work, share API limits, browser and GPU capacity, recover when workers crash, hand work off safely, stop once enough is done, and see what’s blocked and why. Plug it into the agents you already run.









"stop once enough is done" is the feature I'd pay for. The failure mode I hit running agents for AskDeck isn't crashes, it's five workers finishing the same deck three different ways. How does runstate decide "enough", is it a count, a budget, or a signal from the worker?