Auto-Company vs Fusion
A fully autonomous 'AI company' on your own PC: 14 expert-modeled agents ideate, decide, code, deploy and market 24/7 — driven by Claude Code or Codex CLI, with a local dashboard. — versus — A multi-agent software factory: describe a task, agents plan (PROMPT.md), build, review and merge in isolated worktrees — kanban + graph board, missions, agent chat rooms, any model. Early preview.
Both run autonomous 'AI companies' on your hardware: Auto-Company ships 14 expert-modeled agents ideating and shipping 24/7; Fusion makes the company importable and inspectable — org charts, token share per agent, multi-agent chat rooms.
| Auto-Company | Fusion | |
|---|---|---|
| Stars | 1.3k | 964 |
| Forks | 166 | 118 |
| Language | Python | TypeScript |
| License | — | MIT |
| Last activity | 1 months ago | today |
| Topics | agents, orchestration | coding, orchestration |
| Curated connections | 2 | 3 |
Auto-Company — the curator's take
The maximalist experiment: give 14 role-agents a company charter and let them run around the clock on your hardware. As a living demo of agentic workflows it's genuinely instructive — watch where autonomy compounds and where it wanders. Treat the 'without human intervention' framing as aspiration, NOT audit: unattended agents produce unattended mistakes at 24/7 speed, token bills to match, and there's NO license file at review time — check before building on it.
Fusion — the curator's take
The most ambitious entry in the agent-factory wave: visual workflow authoring (plan→execute→review graphs you can edit), per-task oversight levels from observe to autonomous with human gates on merges, a multi-node mesh (fleet on a server, steered from your phone), importable 'agent companies' (440+ pre-built agents), and a Command Center with real fleet telemetry. Genuinely MIT and shipping weekly. When NOT: it wears its 'early preview' badge honestly — breadth currently outruns depth, so expect rough edges; if you want minimal-machinery unattended runs from an issue tracker, contrabass is the leaner tool, and a single interactive session needs none of this.