Dalang enters preview
Our first product is open to a small group of early users. Here is what it is, and what it deliberately is not.
Dalang is a multi-agent AI system delivered as one model. You send a request to a single endpoint, and a coordinated pool of models works out the answer behind it. You get one response, not a pile of options to reconcile.
The point is coordination. A single model does well until a task needs several steps, a change of approach, or a second look. Dalang treats those moments as the normal case and plans for them, so quality holds where one pass would stall.
It fits what you already use. Dalang speaks the OpenAI and Anthropic wire protocols, so you can keep your SDK and change three things: the base URL, the key, and the model string. There is no migration to schedule.
Preview means what it says. Access is limited and some rate limits are conservative, but the interface is stable: one endpoint, one answer, two protocols, and one model string.
If you want in, tell us what you are building. We are letting people in based on fit with the problems Dalang is good at, not a waitlist number.