Core platform — agents, tasks, and runs

Orqestr starts here. The core loop — define an agent, give it a task, watch it run — is working end to end.

What shipped

• Agent builder — create agents with a system prompt, model selection (via OpenRouter), temperature, and token limits. Each agent is a specialized worker with a clear role.
• Task board — kanban-style board for creating and tracking tasks. Tasks move through draft → ready → running → completed as agents pick them up.
• Run execution — tasks execute as runs against OpenRouter. Full input/output logging, token tracking, and cost calculation per run.
• Project scoping — agents and tasks live inside projects. Each project is its own workspace with isolated state.

Under the hood

Runs stream progress via SSE so you can watch agents think in real time. Every run step is stored with its full message history, token count, and cost in microcents.