Fusion

This platform is an open source orchestrator for fleets of coding agents. You describe a task in plain language, a triage agent writes it up as a specification with acceptance criteria, and planning, execution and review agents carry it out inside an isolated git worktree, merging the result once every gate passes. The loop is plan, review, execute, review, with pre merge and post merge gates covering documentation, QA, security, performance and accessibility; a clean run is squash merged automatically so history stays linear. Work is organized hierarchically as missions, milestones, slices, features and tasks, with autopilot progress tracking across hundreds of tasks on one board. A model router picks a provider per task with cheap tier routing and fallback, and the supported set spans Anthropic, OpenAI, Gemini, Z.ai, Ollama, Hermes, Droid, Cursor and any agent speaking ACP, local or hosted. A multi node mesh shards work across a laptop, a Mac mini, a Linux server, a cloud VM or a phone while keeping settings, missions and board state in sync, and a Command Center reports tokens, cost, autonomy ratio, per agent output, model mix and system vitals in real time. The project publishes no price list and has no pricing page. It is open source, runs straight from npm with no install step, and is also offered through a shell installer and a Homebrew tap. Model usage is billed by whichever provider you point it at, so the running cost is your own API spend. The site describes the software as an early preview, shipping daily.