DocumentationK-Nets
K-Nets
Knowledge networks — durable AI workflows built from reusable K-Nodes, which keep running through failures and restarts.
A K-Net (knowledge network) is a workflow that an agent or a person builds from K-Nodes: reusable steps such as "fetch a page", "ask a model", "write a row". K-Nets run durably: each step is retried on failure, and a run survives restarts. Their runs show in Instances.
Pages
| Page | What it is for |
|---|---|
| K-Nets | Your knowledge networks. Manage workflows (Agentic Master Loop, ResearchOrchestrator), toggle between Grid and List views, and search across networks |
| New K-Net | Start from a template (Agentic Master Loop, Durable Resilient Chatbot, KnetDB Agent, KnetKB Agent) or start blank |
| Batch actions | Select networks to Export as JSON, Duplicate, Upload JSON, or Delete |
| Marketplace | Ready-made K-Nets and K-Nodes to add to your workspace |
| K-Nodes | Reusable atomic steps that K-Nets are composed of |
| Flows | Visual flow canvas for low-code node wiring |
| Components | Custom code and integration components |
Typical tasks
- Ask Kue to start one of the Let Kue set it up goals; it creates the K-Net for you.
- Add a ready-made K-Net from the Marketplace and run it on your own data.
- Run a saved K-Net from outside:
POST /api/v1/knet/run/{workspace_id}/{endpoint_name}(set the endpoint name in the canvas under Settings → Endpoint Name).
See Building on the Canvas for the canvas itself.