Roadmap

What’s planned next.

A forward-looking view of where Krypton is headed. Order suggests rough priority; nothing is committed and dates aren’t promises.

Near term

ItemStatusNotes
OpenTelemetry tracingPlannedOTLP exporter across gateway → activator → pod; W3C traceparent propagation
MCP resources + promptsPlannedresources/list, resources/read, prompts/list, prompts/get
Schema-driven MCP tool formsPlannedReplace JSON textarea with auto-generated inputs per tool
Invocation history (Postgres)PlannedSurfaced in UI; queryable via control plane
TCP-dial readiness checkPlannedCloses the kube-proxy programming race on first cold-start
Cert-manager integrationPlannedOne-shot webhook enablement; today requires manual cert plumbing

Medium term

ItemStatusNotes
Serverless mode (GA)PausedCode is functional; needs more end-to-end tuning before recommending
GPU-aware schedulingPlannedspec.gpu: { count, type } → nodeSelector / tolerations; MIG slicing
AI-native observabilityPlannedOTEL GenAI semantic conventions, token usage roll-ups
Per-agent network policiesPlannedGenerated egress / ingress rules from spec
Secure sandbox runtimePlannedSandbox CRD with gVisor / Kata runtime classes for AI coding agents

Longer term

ItemStatusNotes
Multi-cluster federationExploringFederated control plane; cross-cluster routing with failover
Authentication & multi-tenancyExploringOIDC-backed UI/API, per-tenant namespaces with quota
Image provenanceExploringSigstore / cosign signature verification at admission
Sign in with Vercel SSOExploringOAuth into the operator UI

Out of scope

Some things people ask about that we deliberately won’t build:

ItemWhy not
Built-in model hostingKrypton runs agents, not models. Use any provider.
A vector DBAgents bring their own (managed Pinecone, pgvector, Weaviate, …).
Prompt managementBelongs in agent code or a sibling product.
Non-Kubernetes deploymentThe design hard-relies on the Kubernetes API for desired-state, scheduling, and lifecycle.

Have a feature in mind that’s not listed? Open an issue.