AI agent platform
AI Control Center
AI agents that ship work, under human control.
AI Control Center is an agent platform operated from Telegram. A task comes in as plain language, a planner breaks it down, and a real AI runtime such as Claude Code or Codex executes it inside an isolated workspace. The result comes back as a verifiable artifact, never as an unchecked claim. Anything with an external effect, like a merge or a publish, waits for human approval tied to the exact item being approved. It is built for one goal: delegating real engineering and operations work safely and keeping a full audit trail.
- Python
- PostgreSQL
- Claude Code + Codex
- Telegram Bot API
- Docker
What it does
Runtime-agnostic core
Four interchangeable AI runtime adapters run behind one contract, so swapping or adding a model provider does not touch the core.
Human-gated effects
Merges, deploys and publications require an approval bound to the exact artifact. Nothing ships on an implicit yes.
Sealed execution
Agents work in isolated worktrees and containers with restricted outbound network access, and every action is recorded.
Run from your phone
Tasks, plans, approvals and results all flow through Telegram, so work moves forward without opening a laptop.