Roadmap

The vision: an AI-native cloud built from composable primitives — compute, storage, and workflows.

London control plane brought up: multi-tenant compute + storage, networking, security baseline, and deployment automation.

Serverless vector engine — tenant isolation, fast provisioning, scalable indexing and search. Now the engine inside K3 Vector.

Unified authentication and authorization across the whole stack, with end-to-end tenant isolation.

Serverless compute: functions and containers via SDK, Dockerfile, no-Dockerfile build, BYO image, and GitHub deploys.

Typed pipeline DSL — the glue that wires every primitive together. Compile-time validation, resumable threads, streaming pipes.

Pre-built model catalog — OpenAI-compatible Chat & Embeddings, Cohere Rerank, and a generic Predict endpoint.

S3-compatible storage where each bucket carries HTAP Tables and hybrid Vector search — Objects, Tables, Vector as one.

The first developer surface: CLI, Python & Node SDKs, an MCP server, and docs.

The full primitives suite opens to Early Access — Ignite + Ignite Models, K3 Objects / Tables / Vector. Sovereign by design, pay-as-you-go.

Reasoning, MCP tools, and managed actions with budgets, rate limits, and blast-radius guardrails.

EU and UK live today; Middle East and Africa next — keeping your data plane in-region.