Run Agents and Workflows[NEW]
Overview
The run screen in Expert Agent Studio lets you use published agents and workflows without building or changing them. You pick an asset, fill in its input parameters, click Run, and see the output on the same page.
This page is mainly for users with the Platform Consumer Policy: business users, stakeholders and teams who consume AI solutions built by others. Builders can use the same screen to try a published asset the way a consumer would.
With the run screen, you can:
- View all published agents and workflows that are public or shared with you
- Run automation agents, conversational agents and workflows
- Upload files or enter values as inputs, and view the output after the run
- Create and manage your own credentials when an asset needs them
- View all of your transactions across automation agents, conversational agents and workflows
What you cannot do as a Platform Consumer:
- Build, edit or delete agents, workflows or other assets
- View an asset's configuration, details side panel or benchmark options
- View the FinOps dashboard in Governance
Before You Begin
You need three things in place before you can run an asset: the right policy, a shared asset, and any credentials the asset uses.
- Policy: your Tenant Admin must assign you the Platform Consumer Policy, or another policy that allows running assets, in at least one workspace.
- Asset access: the agent or workflow must be published, and either public in your workspace or shared with you. Unpublished or private assets do not appear in your list.
- Credentials: if the asset connects to an external system, it may need a credential. You can create one ahead of time in the Credentials module, or create it when the platform shows Authentication Required during a run.
Modules you will see with the Platform Consumer Policy
If an asset you expect is missing, ask its owner to publish it and share it with you, or ask your Workspace Admin to check your policy.
How To: Run a Workflow or Agent
This example runs an automated workflow called invoice extraction. The same steps apply to automation agents.
Step 1: Open Expert Agent Studio
- Log in to Purple Fabric. Expert Agent Studio opens by default for Platform Consumer users.
- The page lists the published agents and workflows available to you. Use search and filters to find an asset.
- Workflows appear in the left-side panel in the same way as automation agents.
Step 2: Select the asset
- Click the workflow or agent you want to run, for example invoice extraction.
- The run screen opens. The asset name appears at the top, with a tag showing its type (for example, Workflow).
- Click the back arrow next to the name to return to the asset list at any time.
Step 3: Fill in the Input Parameters
The Input Parameters panel shows the fields the asset needs. The fields are different for each asset, because its builder defines them.
- For a file field (for example, Invoice Docs), do one of the following:
- Click Browse and choose a file from your computer.
- Enter a file path or URL in the field.
- For a text field (for example, Organisation Name), type the value.
- Fill in every field the asset needs before you run it.
Step 4: Add credentials, if prompted
If the asset uses a credential that is not mapped, the platform shows the Authentication Required window before the run starts.
- In Select account, choose a credential under Personal or Shared.
- If no account fits, create one under If you do not find the appropriate account, please link your account. A credential you create here is stored as your Personal credential, visible only to you.
- Click Continue to proceed to the run.
Step 5: Run the asset
- Click Run.
- Wait for the run to finish. The output appears in the Output panel below the inputs.
Step 6: Review the output and transaction
- Read the result in the Output panel. Until you run the asset, the panel shows Output will appear here after run.
- Each run is saved as a transaction. You can find it later in your transaction history (see
Event-triggered Workflows, Transactions and Credentials ).
Running assets from Governance: You can also run automation agents and workflows from the Create Transaction tab in the Governance module.
Input Parameters and Output
Every field on the run screen comes from the asset's input schema, which its builder set up in Agent Designer or Flow Designer. You only fill them in; you cannot add or change fields.
Screen elements
Supported file formats. File fields accept PDF, DOC, DOCX, TXT, CSV, JSON, XML, PNG and JPG files. The maximum file size is 10 MB.
Tips for good results
- Use clear, readable files; scanned documents with low quality can reduce extraction accuracy.
- Check that text values match what the asset expects, for example the exact organisation name.
- Only submit data your organization allows you to process with AI assets.
Event-triggered Workflows, Transactions and Credentials
Event-triggered workflows. Some workflows start automatically when an event happens, such as a new file or message arriving, rather than from a manual run. When you click one of these workflows, the platform shows a message that it is event-triggered, along with details about its trigger. You cannot start it with Run.
Your transactions. Every run you start is recorded as a transaction. You can view all of your own transactions across automation agents, conversational agents and workflows. You cannot see transactions started by other users.
Your credentials. Credentials are managed in the Credentials module. See Permissions by Policy for the full rules.
Sensitive values such as tokens and keys are always masked.
Limitations
- Only published assets that are public or shared with you appear in your list.
- You cannot view an asset's details side panel, configuration or benchmark options.
- You cannot build, edit, delete or publish assets.
- In Governance, you see agent listing details only, not the FinOps dashboard.
- Event-triggered workflows cannot be run manually.
- File inputs are limited to 10 MB.
FAQs and Troubleshooting
FAQs
1. Can I change how an agent or workflow works?
No. The Platform Consumer Policy is read-only. Contact the asset's owner for changes.
2. Why can't I see the side panel or configuration of an agent?
Consumers only run assets. Asset details, configuration and benchmark options are hidden by design.
3. Can I run conversational agents too?
Yes. You can start transactions with automation agents, conversational agents and workflows that are published and shared with you.
4. Who can see my personal credentials?
Only you. Personal credentials are never visible to other users.
5. Can I see other people's transactions?
No. You see only the transactions you started.