Krea's Node Agent builds workflows from prompts
Krea has added an agent to its Nodes canvas that turns a one-sentence request into a runnable generative pipeline, moving agentic workflow-building from developer tools into creative production.
Krea described a Node Agent built into Nodes, its node-based tool for creative work. You type what you want, for example "cinematic video of a futuristic city at sunset". The agent then plans a pipeline, places and connects the nodes, and runs the job. Krea says this is a separate product from Krea Agent, its chat-based production agent.
The agent first reads the existing canvas, including nodes, connections and earlier outputs. It reuses what is already there, such as a style node from a previous session, rather than starting from scratch. Before it builds anything, it shows a plan. In Krea's example the plan has four stages: prompt refinement, image generation on K1 or Flux, enhancement, then video. Users can swap models or drop stages before approving it.
Once the graph is built, the agent checks it before running. It fills in missing parameters. Where an output feeds an incompatible input, it repairs the connection or adds a conversion node. It then shows the compute cost for each node, and nothing runs until the user confirms. When a user edits the graph, only the nodes downstream of the change run again. Upstream results stay cached.
Krea says planning and reasoning come from frontier LLMs, including Anthropic's models. Generation runs on Krea's own models and others on the platform, including K1, Flux and Imagen.
The post gives no pricing, usage numbers or details about availability. It also doesn't say which LLMs are used besides Anthropic's.