Introducing VERSUR Agent

Building a workflow by hand means knowing the palette, the wires, and the field names before you know whether the idea works. Most of the time you know the outcome - not the graph.

Today we are introducing VERSUR Agent - the first-class workflow assistant for VERSUR. Build, ask, edit, and optimize - and debug from real runs. Describe what you need in plain language. It plans the graph, places real blocks, and defines how they connect and run - then leaves a canvas diff you can Preview, Accept, or Reject.

Build

You ask for an outcome - a brief-to-render pipeline, a site analysis graph, a review loop with parallel options. VERSUR Agent plans the steps, then builds against your real block catalog. Blocks can appear on the canvas one by one while it works, so you are not staring at an empty graph until the end.

• Real blocks from your palette - not a separate description layer.
• Connections between blocks, with inputs and outputs wired to the same paths you use by hand.
• Skills, tools, loops, and loop playbooks configured on the Agent and design-agent steps it creates.
• A canvas diff for new and edited blocks until you Accept or Reject.
• Optional test run after a build - Run or Skip - so it can read failures and keep editing.

Ask

VERSUR Agent is also how you understand the graph. Ask what the workflow does, how data flows between blocks, which output path to wire, or how a block should be configured. It reads your canvas, the block catalog, best practices, and in-app docs - and can pull workflow examples when you want a pattern to start from.

Edit and optimize

Ask it to add a branch, swap a model, insert a design agent, retarget an input, or fix a broken wire. It edits the same workflow in place - connections, fields, skills, tools, and loops included. Changes stay provisional until you Accept or Reject.

Optimize is part of the same job. Ask why a run is expensive or how to improve performance - it can surface duplicate model calls, fat references, and leaner wiring, then apply the edits when you want them.

Debug

After a build or edit, VERSUR Agent can offer a test run. If a block fails, it reads the failure from the run console, edits the graph, and can offer another run - without starting over from a blank canvas.

Builder vs agents on the canvas

VERSUR Agent builds and maintains the workflow - including how each step is configured. The Agent block and design agents it places are the reasoning steps on that graph: skills, manuals, tools, loops, Brain, Bridge-backed local models. One is the workshop for the graph. The others are what runs inside it.

Use VERSUR Agent when you need the pipeline. Use Agent blocks and design agents when you are shaping a single studio-grade reasoning step by hand.

How to start

• Open a workflow and start VERSUR Agent.
• Describe an outcome, ask a question about the graph, or ask it to change, optimize, or fix something.
• Watch blocks appear or update; accept the diff when the layout looks right.
• If it offers a test run, tap Run to verify - then ask it to debug anything that failed.

Available now

VERSUR Agent is available in the workspace today. Build, ask, edit, optimize, and debug - the canvas is the deliverable.

Brain is where the practice remembers. Design agents are how the practice designs. Bridge is how inference can stay on your machine. VERSUR Agent is how the workflow gets built and kept working.

Design Agents · Bridge · Brain · Changelog