Engineering deep dives, industry perspective, implementation notes, and practical recipes.
Search covers titles, summaries, tags and full article text.
How KDCube governs MCP in both directions: per-agent consumer inventories, app-native provider surfaces, delegated authority, dual-era wire support, named services, and economics around paid work.
How KDCube apps move state: the durable Data Bus for inbound changes, the live relay streaming to watchers, and the scene bridge into embedded widgets.
No magic token table: handles, secrets, metadata, edges, and sessions each live where their lifecycle fits. The storage model is the design.
One app consumes a narrowly allow-listed MCP service, exposes its own governed tools, and hands automations expiring access — credentials never travel.
What you actually write for a KDCube app: four starting points, the runtime services around each, and the point where the work stays yours.
The platform renders catalog pages; until now an application could recolor exactly three things. This entry records the presentation contract that ends that: every visual decision in the renderer reads a --kdcpub-* design token, and an application overrides tokens or loads its own stylesheets — declared in its config, per alias and per fold. The platform ...
AI agent governance from policy to runtime enforcement: identity, authority, action policy, isolation, cost controls, and evidence working as one system.
AI agent infrastructure beyond the agent loop: how a self-hosted runtime adds identity, cost controls, isolation, and deployment around your framework.
A working agent is product IP, not migration debt. Preserve its LangGraph core and add a bounded async host layer for ordered turns, streaming, user-bound state, and reload.
In KDCube's reference split-Docker profile, generated code runs in a separate networkless executor while approved tools act through a trusted supervisor.
Every integration wants to act as someone, somewhere. One broker keeps identities, claims, grants, and secrets distinct across three directions of trust.
KDCube counts LLM calls, embeddings, and web searches as usage — recorded, priced, and settled against a user's budget. Your application's own paid work can join them. A decorator, four small extractors, and one usage object turn a metered API call into first-class usage the economics model can charge.