Allow input files in published agent

Question

https://learn.microsoft.com/en-us/microsoft-copilot-studio/image-input-analysis

Hi there,

We want to deploy our agent and make it able to read screenshots put in by end-users. However, it seems to only work in the test environment. Can you help with this, I can't find any relevant q&A's on the website. It seems very weird to me that in a test environment it would work and in the published teams version it doesn't work. The whole point of testing is, is such that it works with the end-user.

Hope you can help me,

Thanks!

Answer 1

In Copilot Studio, file and image upload support depends on both the agent’s settings and the channel where it’s published. What is seen in the “Test your agent” pane is the full capability of the agent, but not every channel exposes those same capabilities to end users.

Key points from the product behavior:

1. File uploads must be enabled in the agent settings
   • In the agent’s Settings → Generative AI → File processing capabilities, File uploads must be turned on.
   • After enabling, test again in Test your agent using the paper-clip icon to confirm screenshots/images are accepted and analyzed.
2. Channel limitations apply after publishing
   • Not all channels support user file uploads, even if the agent does.
   • For example, when an agent is published to a SharePoint channel, “users can't upload files.” In that case, uploads work in the test canvas but not for end users in SharePoint.
   • Similar limitations can exist per channel (Teams, web, etc.), depending on how that channel surfaces attachments.
3. Teams-specific behavior
   • For Teams-based conversations, files can be present as attachments in the conversation and accessed via System.Activity.Attachments or via a Question node configured to request a File.
   • To make the agent explicitly ask for a file (such as a screenshot) and process it:
     1. Add a Question node in a topic.
     2. Under Identify, select File.
     3. In Question properties → Entity recognition, select Include file metadata.
   • Alternatively, use a Set variable value node with First(System.Activity.Attachments) to pick up a file already attached in the Teams message.
4. Supported file types and limits
   • For screenshots and images, supported types include JPG, PNG, WebP, nonanimated GIF.
   • Ensure the images are within the size limit (15 MB per file) and are clear enough for analysis.
5. Content moderation and environment constraints
   • If content moderation flags the image, the agent won’t respond; restarting the conversation may be required.
   • If the agent is in a customer managed key enabled environment, “adding files as input is allowed, but the agent doesn't process the files.” In that case, uploads appear but won’t be analyzed.

Practical steps to troubleshoot the difference between test and published behavior:

1. Confirm File uploads is enabled in Settings → Generative AI → File processing capabilities and saved.
2. Verify the agent is not in a customer managed key enabled environment if file processing is required.
3. Check the channel:
   • If using SharePoint, file uploads from users are not supported.
   • If using Teams, ensure users are either:
     • Uploading/attaching the screenshot to the message, and the topic uses System.Activity.Attachments, or
     • Responding to a Question node configured with Identify = File.
4. Re-publish the agent after any configuration changes and test again in the target channel.

If the test canvas works but the published Teams experience does not, the most common causes are channel limitations, missing file-question configuration in the topic, or environment restrictions (such as CMK) preventing processing.

---

References:

• Allow file input from users
• Pass files to agent flows, connectors, and tools
• Unstructured data as a knowledge source
• File attachments in Azure SRE Agent
• Tutorial: Share files and screenshots in Azure SRE Agent