Build and change the canvas, and publish the version the team runs.
One person builds.The whole team runs it.
App Mode turns any workflow into a simple form. Teammates pick the product, press Run and get on-brand work. The canvas stays with the builder.
Pick a product and press Run.
Everyone stays in their lane.
Complexity stays with the people who build. Everyone else gets a tool that works the same way every time.
Run the app, fill in the form, get the results. They never see a node.
When the workflow has one, every run waits for a person to approve, reject or edit before anything ships.
Why teams switch.
The person who knows the models builds the workflow once. Everyone else gets the result without the learning curve.
Teammates fill in a form. The prompts, models and brand rules are already inside the workflow you built.
Everyone runs the version you deployed, so output looks the same on Monday and on Friday, whoever presses Run.
The credit cost shows before Run is pressed. Nobody finds out what a batch cost after it has spent.
The form, your own code over the API, or Claude and ChatGPT. One workflow behind all of them.
Apps teams run.
Any workflow your team repeats with different inputs can be an app. A few shapes we see most.

Merch teamNew product launch
Packshot + market A set of ads in every channel size

Regional marketersLocalize a winner
Winning ad + language The same ad, re-voiced for their market

Performance teamHook tester
Product + angle UGC-style video hooks to test this week

AgentsListing to tour
Listing photos A twilight hero and a walkthrough clip

Agency account managersClient deliverable
Client + offer That client's ads in their own brand kit

App marketersStore to ads
Store screenshots Install ads for each network
Edits don’t leak. Deploys do.
Rework the canvas as much as you like. The team stays on the version you approved until you deploy the next one.
Nothing ships unseen.
Wire a review step into the workflow and every app run waits there for a person. Guardrails live in the graph, so they apply to every run.



The run is waiting for your review.
Run it from anywhere.
One app, three ways in. The team uses the form, your systems call the API, and your assistant runs it from a chat.
For the team
A link in the workspace. Fill in the fields, see the cost, press Run, watch it work and open earlier runs any time. Nobody needs to know a node exists.
LanguageGermanAbout 70 creditsRunWho sees what.
Pick how far the app travels. Whoever opens it gets a form and results, never the canvas behind them.
Only you. Good for testing the form before anyone else sees it.
Everyone in your workspace can open and run it.
Anyone with the link, inside your company or not.
Listed in the public app directory for anyone to find.
They see
- The fields you chose, with your labels
- Live progress while it runs
- The outputs you picked: images, video, text, audio or files
- Their earlier runs, ready to rerun
They don’t
- The node graph
- Brand and context nodes, which still shape every result
- Your drafts, until you deploy them
Before you ask.
Do I have to design the form?
No. You pick which input nodes become fields: a text prompt becomes a text box, an upload node becomes a file picker. Other inputs and brand nodes stay behind the scenes.
What happens when I change the workflow?
Nothing, for the team, until you deploy. Publishing pins a version, so they keep running the one you approved.
Can a run wait for approval?
Yes. Add a human-review node and every app run pauses there until someone approves, rejects or edits it.
Will a teammate burn the budget by accident?
Every run shows its credit cost before Run is pressed.
Can people outside my workspace use it?
If you want them to. Share it by link, or list it in the public app directory. Keep it private or workspace-only otherwise.
Can we run an app from our own code?
Yes. Every deployed app has an API endpoint. An owner or admin creates an API key, you post the inputs, then check the run until it finishes and read the outputs.
Can I run an app from Claude or ChatGPT?
Yes. With Orisu connected, ask your assistant to run the app. It fills in the fields, runs it and shows the results in the chat.
What kinds of workflows make good apps?
Anything your team repeats with different inputs: a product launch set, localizing a winning ad, hooks to test, client deliverables, listing videos.
Can someone go back to an earlier result?
Yes. App users can rerun with new inputs and open their earlier runs.
Can an app run over a whole catalog?
Yes. Mount it as a column in Feed and it runs once per product row. Apps in Explore have a Run over a feed button.
Can I send files to an app from code?
Yes. Upload the file to the assets endpoint, up to 100 MB, and pass the asset it returns to the app's upload field.
What happens to the app if I delete the workflow?
The app goes with it. Deleting an agent also deletes its app, so keep the workflow as long as the team uses the app.