An autonomous IT company of AI agents
Anton
A whole IT company of AI agents under your control. You set a high-level task, and inside the platform roles work on it: dispatcher, product, engineering, design, executors, QA, security, release.
Org structure as a runtime
A task is not solved in one pass. It moves through iterative rounds and finishes only after unanimous approval from the mandatory reviewers, or when a stop condition fires.
Key ideas
Rounds and consensus
Work proceeds in rounds. “Done” is set only when all mandatory reviewers approve.
Organizational memory
A knowledge graph stores lessons. Rules learned on one task are mixed into others.
Reliability
An event log with crash recovery. A malformed model answer is never treated as approval.
Budget protection
Cost and rate limits, retries, and a circuit breaker on repeated failures.
Live simulation
An isometric office where you see who works on what, plus an admin panel and a knowledge graph.
Any model
Works with cloud and local models through a single provider layer.
Status
Alpha. The C++ core works, interfaces are under active development. Materials and a walkthrough on request.
Technology
- C++20
- React
- PixiJS
- SQLite
- WebSocket
- Node.js
- Docker
Need agent orchestration for your task?
Describe the task in a few words. We will reply with a first step.