← MRM/Projects/AI Company
AI / Agents Experimental Python · Rust · local or cloud models

AI Company

A team of AI agents that each know their job — and remember it.

Single-purpose AI tools fragment knowledge and forget everything between prompts. AI Company is a persistent, multi-specialist team: a CEO agent routes each task to the right department, and every department keeps its own memory. Point it at a local or cloud model and it runs as a company.

company — dispatch
$ company "launch a 2D platformer prototype" CEO → routing to game_dev_guru ├─ assetforge → sprites + tiles generated ├─ genforge → SFX + ambience generated └─ scaffolds a playable build 28 departments · shared persistent memory

Representative output — not a live instance.

What it does
28 specialized departments — content/SEO, bookkeeping, invoicing, lead-gen, sales, proposals, legal review, grant writing & discovery, cyber-defense, game development, and more.
CEO orchestration — a dispatcher classifies each task and routes it to the best department, or you can target one directly.
Persistent memory — each agent keeps episodic, semantic and procedural memory in a local store, so the "company" doesn't start from scratch each query.
Pluggable backend interface — Ollama cloud and local are implemented today; Claude, OpenAI and a local RWKV model are planned.
It builds things — the game department coordinates AssetForge and GenForge to scaffold a playable game end-to-end.
Proof
0
agent departments
stateful
persistent per-agent memory
2
live backends · Ollama cloud/local
tested
real game-build workflows
PythonRust (CLI + server)SQLite memoryOllama (cloud/local) · others planned
🔒
Architecture shown; needs a model backend to run. Source, agent prompts and orchestration are private — this is a capability showcase, not a distribution.

Want an AI team,
not just a chatbot?

I build multi-agent systems that route work, keep context, and actually produce deliverables. Tell me the workflow you'd hand a team.

Start a project →