Beta Release Guide
AI Agent for Boards
Everything you need to get started and get the most out of your agent.
What is The Agent?
The AI agent lives directly on your board and can do anything you can do — generating images, chaining modules, running workflows — without any of the manual effort. Instead of configuring modules and writing prompts yourself, you describe the end result you want and the agent figures out the rest autonomously.
Beta release: All agent generations are credit-free for the full duration of the beta. Experiment freely, push the agent hard, and don't hold back — your usage and feedback directly drives the roadmap.
What the Agent Takes Off Your Plate
The agent removes the two hardest parts of working on the platform:
Module Confusion: You no longer need to figure out which modules to use or in what sequence. The agent selects and chains the right modules automatically based on the outcome you describe.
Prompt Construction: Writing effective prompts is a learned skill that slows down new and experienced users alike. The agent crafts optimized prompts behind the scenes, tailored to your specific goal — not a generic template.
Where to Find It and How to Get Started
The agent lives on your board. Open the agent panel from the board toolbar, describe the end result you're looking for in plain language, and it gets to work.
Starting Your First Task
Open the agent panel from the board toolbar
Type a plain-language description of what you want — not a technical prompt, just the goal
The agent confirms its plan, then starts working
You can keep working on your board, close it, or walk away — the agent continues running in the background
You can leave or close the board at any time. The agent keeps generating and your outputs will be ready when you return.
Board title, Share, your profile, and canvas toolbar
Tools, Assets, History, Settings
Your workspace where you arrange nodes and run workflows
Running Multiple Tasks At Once
You can run up to 10 agent tasks concurrently. You don't need to wait for one to finish before starting another.
Example: Parallel Work
Agent task 1 → Running a 12-color palette variation workflow across 4 base designs
Agent task 2 → Rendering a full sketch-to-on-body pipeline across multiple poses
You → Editing your mood board or starting a new project — completely uninterrupted
Kick off tasks before a meeting, before lunch, or before bed. Come back to review a batch of finished outputs.
Tracking Your Tasks
Open the history dropdown to see the status of all your agent conversations at a glance:
Pending your response — the agent has asked you something and is waiting
Generating — the agent is actively working
Ready to review — outputs are complete
Supported Workflows (Day One of Beta)
Color variations using a palette — apply color variations across designs without masking
Sketch to on-body — full pipeline from sketch to styled on-body renders across multiple poses, in a single prompt
Batch sketch to render — render multiple sketches in one go
Print recolors applied to models — recolor prints and apply them to model shots
Any chained combination of the above — from sketch to render to on-body in a single prompt
Shared Boards and Multi-Agent Collaboration
On a shared board, each collaborator can join with their own agent. Multiple agents from different users can work on the same board simultaneously — each running tasks independently while all outputs land in the shared space.
This means your whole team can use the agent on the same board at the same time, with each person running their own parallel workflows.
Berry Personalities
When you use the agent, you are randomly assigned one of 10 berry personalities. Each berry has its own distinct character and style.
What It Does
Your berry personality adds small, expressive notes as you type and as generations happen — a little companion that reacts to what you're doing. It doesn't change the tone of the actual chat or affect any outputs. It's purely a fun, ambient layer on top.
Meet the Berries
The mysterious one. Always overdressed, never explains why.
Sweet, soft, and secretly the most powerful one in the room.
Perpetually snacking. Zero chill. Maximum energy.
Has opinions about your color palette. Very strong opinions.
Don't talk to her when she has her headphones on. She's in the zone.
Has read every book. Will recommend five more you haven't heard of.
Gets 47 ideas before breakfast. Writes none of them down.
Manifests things into existence. No one knows how.
On their third sketchbook this month. Erasers are their love language.
Unbothered. Moisturized. Sipping slowly. Always right.
Changing Your Berry
Your berry personality adds small, expressive notes as you type and as generations happen — a little companion that reacts to what you're doing. It doesn't change the tone of the actual chat or affect any outputs. It's purely a fun, ambient layer on top.
Best Practices
Describe Outcomes, Not Steps: Don't tell the agent which modules to use or how to structure a prompt. Just describe what you want the final result to look like. "Create on-body renders of these 6 sketches in 3 poses each, editorial style" is better than trying to configure each step manually. Let us know if the agent does anything unexpected when chaining together modules — your feedback shapes what we fix first.
Be Specific About the End Result: The more context you give about the desired output — mood, style, quantity, variation type — the better the agent can make decisions on your behalf.
Use Parallel Tasks for Batch Work: Don't run workflows sequentially when you can run them in parallel. Kick off your color variation task and your sketch-to-on-body task at the same time, then review both when they're done.
Let it Run Overnight: The agent is persistent — close the board and come back the next day. This is especially useful for large batch jobs or multi-step workflows across a full collection.
Iterate in the Same Conversation: Each agent chat is a context-aware thread. If you want to refine an output, continue in the same conversation — the agent knows what it already generated and can adjust from there.
Check the History Dropdown: Get in the habit of checking the history dropdown to monitor which tasks need your input (pending) versus which are running or done. This keeps you from missing a task that's waiting on a decision from you.
How the Agent Learns Over Time
The agent isn't static. Over time it learns and evolves based on how you work.
Individual Personalization
The agent learns your preferred render styles and finishes, how you like to sequence workflows, the kinds of outputs you approve vs. refine, and your personal shorthand — getting faster and more accurate the more you use it.
Organization Personalization
At the org level, the agent learns your brand's aesthetic and visual language, your signature color systems and palettes, recurring workflow patterns across your team, and what "good" looks like for your organization's output standards.
Longer term, the agent evolves to feel custom-built for your team — not a generic tool, but one that understands how your organization creates.
This is one of the biggest reasons to join the beta early. The sooner you start using it, the sooner it starts learning your style. Beta participants get a meaningful head start on personalization.
How to Leave Feedback
Your feedback during the beta is the most valuable input we have. Here's how to share it:
In-conversation feedback — use the thumbs up / thumbs down reactions on any agent output directly in the chat
Bug reports and specific issues — flag these in the dedicated beta feedback channel
Workflow suggestions — if there's a workflow or use case you want supported, note it in the feedback channel
Berry feedback counts too — if a personality feels off or you love one in particular, let us know
We're monitoring feedback daily during the beta. Nothing goes into a void.
