Engineering deep dives, industry perspective, implementation notes, and practical recipes.
Search covers titles, summaries, tags and full article text.
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 ...
Effective AI agent governance connects a declared policy to the exact place where identity, authority, money, code, data, and side effects become real. Output controls, action policy, observability, and runtime boundaries are complementary parts of that system.
Keep the agent, tools, user experience, and product you already built. KDCube gives them an open-source, self-hosted place to operate: deploy from Git, serve through APIs, webhooks, MCP, or streamed conversations, and add identity, storage, files, isolated execution, integrations, and cost controls when the product needs them.
You have a capable LangGraph agent — a multi-node graph, nested subagents, its own retrieval and memory, its own persistence — running on one machine for one user. KDCube hosts it at scale without rewriting it. Your graph definition and behavior stay yours; KDCube builds a fresh graph instance for each bound turn and discards it afterward. The platform ad...
An agent can write useful code: inspect data, join records, transform files, build a chart, or assemble a report. That code needs CPU, libraries, and a workspace. Network access, provider credentials, deployment secrets, and the authority to change an external system belong somewhere else. KDCube separates those jobs. Generated code computes inside a boun...
Every integration wants to act as someone, somewhere. Send mail through the user's Gmail account. Let a script call an application on the user's behalf. Accept a Telegram message and connect its sender to the platform user who owns the work. Each journey carries a different proof, credential, and direction of authority. Connection Hub gives them one gover...
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.
Bring your own agent — your graph, your framework, your control loop. KDCube hands it user messages through one door, off the event bus: ordered, one turn at a time per conversation, exactly once, nothing lost. A run-to-completion loop gets that for free; a loop that can absorb input mid-flight opts into more.
Your application already knows how to spend money on a user's behalf — an LLM call, an embedding, a web search. The question is whether the platform gets to decide, before the call runs, that this user can afford it, and to charge the real cost afterward. KDCube makes that decision the same on every surface, without your application reimplementing plans, ...
A KDCube tool can call your service as the user — not with a shared system key, not with a platform login token, and not with a secret copied into code. The user connects their account once through Connection Hub; after that, tools and named services resolve the user's provider token only when the requested claim is approved.
An application already owns a browser experience: source files, a build, a main view, APIs, and the platform session around them. It should not need a second web server, a hand-written reverse-proxy block, or a separate authentication implementation to become a website. Declare the site in the application; KDCube builds it, routes it, and gives a CDN one ...
A tool schema tells an agent how to call a function. It does not tell the runtime whether that call reads, writes, or can safely begin before the model has finished generating the rest of its round. KDCube tool traits add that missing policy layer — metadata the runtime reads and enforces , not prompt advice.