xuiltul / animaworks

AnimaWorks — Organization-as-Code **No one can do anything alone. So I built an organization.** A framework that treats AI agents not as tools, but as autonomous people. Each Anima has its own name, personality, memory, and schedule. They talk to each other through messages, make their own decisions, and work as a team. Just talk to the leader — the rest takes care of itself. Workspace dashboard: each Anima's role, status, and recent actions displayed in real time. 3D office: Animas sitting at desks, walking around, exchanging messages — all on their own. **日本語版 README** | **简体中文 README** | **한국어 README** --- How It Compares | | AnimaWorks | CrewAI | LangGraph | OpenClaw | OpenAI Agents | |--|-----------|--------|-----------|----------|---------------| | **Design** | Autonomous org | Role-based crews | Graph workflows | Personal assistant | Lightweight SDK | | **Memory** | Brain-inspired: consolidation, 3-stage forgetting, 6-channel priming with trust tags | Cognitive memory with manual forget | Checkpoint + cross-thread store | SuperMemory knowledge graph | Session-scoped | | **Autonomy** | Heartbeat (observe/plan/reflect) + Cron + TaskExec — runs 24/7 | Human-triggered | Human-triggered | Cron + heartbeat | Human-triggered | | **Org structure** | Supervisor→subordinate hierarchy, delegation, audit, dashboard | Flat roles in a crew | — | Single agent | Handoffs only | | **Process model** | One OS process per agent, IPC, auto-restart | Shared process | Shared process | Single process | Shared process | | **Multi-model** | 6 engines: Claude SDK / Codex / Cursor Agent / Gemini CLI / LiteLLM / Assisted | LiteLLM | LangChain models | OpenAI-compatible | OpenAI-focused | > AnimaWorks is not a task runner — it's an organization that thinks, remembers, forgets, and grows. It supports your business as a team and can be operated as a company. --- :rocket: Try It Now — Docker Demo 60 seconds. Just an API key and Docker. A 3-person team (manager + engineer + coordinator) starts working immediately, with 3 days of activity history pre-loaded. Read more about the demo → > Switch language/style: — see all presets --- Quick Start macOS / Linux / WSL: Windows (PowerShell): If you want to use OpenAI Codex without an API key, run before the first launch. Open **http://localhost:18500/** — the setup wizard walks you through it: • **Language** — pick your UI language • **User info** — create your owner account • **Provider auth** — enter an API key, or choose Codex Login for OpenAI • **First Anima** — name your first agent No editing needed. The wizard saves everything to automatically. The setup script installs uv, clones the repo, and downloads Python 3.12+ with all dependencies. It covers **macOS, Linux, and WSL** with no pre-installed Python required. On **Windows**, use the PowerShell/manual steps above. > **Want to use a different LLM?** AnimaWorks supports Claude, GPT, Gemini, local models, and more. Enter your API key in the setup wizard, or use **Codex Login** for OpenAI/Codex. You can change it later from **Settings** in the dashboard. See API Key Reference below. Alternative: inspect the script before running If you'd rather review the script before executing it: Alternative: manual install with uv (step by step) Alternative: manual install with pip > **macOS users:** The system Python ( ) is 3.9 on macOS Sonoma and earlier — too old for AnimaWorks (requires 3.12+). Install via Homebrew ( ) or use the uv method above, which handles Python automatically. Requires Python 3.12+ already on your system. --- What You Get Dashboard Dashboard: 19 Animas across 4 hierarchy levels, all running with real-time status. • **Chat** — Talk to any Anima in real time. Streaming responses, image attachments, multi-thread conversations, full history • **Voice Chat** — Talk with your voice right in the browser (push-to-talk or hands-free). Supports VOICEVOX / SBV2 / ElevenLabs • **Board** — Slack-style shared channels where Animas discuss and coordinate on their own • **Activity** — Real-time feed of everything happening across the organization • **Memory** — Peek into what each Anima remembers — episodes, knowledge, procedures • **3D Workspace** — Watch your Animas work in a 3D office • **i18n** — 17 languages for UI; templates in Japanese + English with automatic fallback Build a Team and Let It Run Just tell the leader who you need — they'll figure out the right roles, personalities, and reporting structure, then create new members. No config files. No CLI commands. The organization grows through conversation. Once the team is in place, it runs on its own without you: • **Heartbeats** — Each Anima periodically checks the situation and decides what to do next • **Cron jobs** — Daily reports, weekly summaries, monitoring — scheduled per Anima • **Task delegation** — Managers assign work to subordinates, track progress, and receive reports • **Parallel task execution** — Submit multiple tasks at once; dependencies are resolved and independent tasks run concurrently • **Night consolidation** — Daytime episodes are distilled into knowledge while they sleep • **Team coordination** — Shared channels and DMs keep everyone in sync automatically Memory System Traditional AI agents only remember what fits in the context window. AnimaWorks agents have persistent memory — they search and recall on their own when they need to. Like pulling a book off a shelf. • **Priming** — When a message arrives, 6 parallel searches fire automatically: sender profile, recent activity, related knowledge, skills, pending tasks, past episodes. Agents remember without being told to • **Consolidation** — Every night, the day's episodes are distilled into knowledge — the same mechanism as sleep-time memory consolidation in neuroscience. Resolved issues automatically become procedures • **Forgetting** — Unused memories gradually fade through 3 stages: marking, merging, archival. Important procedures and sk…