Roadmap
What's shipping after v0.1.0 — client SDKs, a dashboard, metrics export, multi-modal inference, a model-driven loop, and an optional managed cluster layer.
v0.1.0 is the runtime core. These are the open-source investments that follow. They are directional, not dated.
On the way
- TypeScript & Python client SDKs over the gRPC gateway — drive the runtime from your language of choice.
- A dashboard — observe every step the runtime takes, surfaced visually.
- Prometheus / OpenTelemetry metrics export — wire runtime telemetry into your existing observability stack.
- Audio multi-modal inference — extend local inference beyond text and images.
- A model-driven dynamic loop — the model computes the topology decision (how to fan out / recover), instead of a fixed recipe shape.
- An opt-in cluster layer for managed multi-node — the hosted side of the coordinator/worker model.
Phases
The runtime (phase 1) is what ships today. Platform & orchestration — multi-agent workflow definitions, a run-graph compiler, and the dashboard — is the next sequenced phase. The long-horizon AI Operating System grows out of the runtime and the platform; it is explicitly a long arc, not a near-term deliverable.
Built in the open
The runtime is Apache-2.0 and developed in public. The managed cloud (hosted tier) is on the waitlist; enterprise and services are contact-sales.