Built this because existing agent frameworks kept surprising me with costs and doing things without asking. And even when they worked, the setup was brutal. Hours of config, dependency hell, and half the integrations broken out of the box just to get a single agent doing something useful. Open Swarm gives you per agent cost tracking, human in the loop approvals, and 4000+ MCP tool integrations that actually work out of the box. Zero config. Works with Claude and expanding to every model (including local). You should be building with agents in minutes, not fighting your framework for days. Would love feedback from the HN community.