What tool do you recommend for my Microsoft Foundry agent to reliably (that's key) store a data file and provide a download link?

Question

What tool do you recommend for my Microsoft Foundry agent to reliably (that's key) store a data file and provide a download link?

Michael Barnett 40

We have had problems using an OpenAPI tool in Microsoft Foundry to reliably (that's key) store a data file and provide a download link. Specifically, when multiple calls are made to the tool, some calls do not complete, return an 'in_progress' status, causing the agent to fabricate non-working download links. This occurs even with this system prompt (added to the existing prompts):

What do you recommend for a more robust tool and approach?

4. **Multiple Files - SEQUENTIAL CALLING REQUIRED**
   • **CRITICAL**: When generating multiple CSV files, you MUST call the SaveCSVFile function **ONE AT A TIME** in sequence
   • **DO NOT make concurrent/parallel function calls** - this causes timeout errors
   • **WAIT for each function call to complete successfully** before starting the next call
   • **Workflow for multiple files**:
     1. Call SaveCSVFile for first file
     2. Wait for response with status "success": true
     3. Store the download link from the response
     4. Only then call SaveCSVFile for second file
     5. Repeat until all files are generated
   • Present all valid download links together after all files are stored
   • If any individual call fails, report which file failed and provide the successful links for others

Michael Barnett 40 Reputation points

2026-01-08T17:30:17.0366667+00:00

We're using the New Foundry UI, but if there are configurations in the old UI, please provide details.
Michael Barnett 40 Reputation points

2026-01-09T15:24:56.8233333+00:00

Thanks Anshika Varshney; however, this response (generated by an LLM?) is non-specific. Can you please address the specific question about a tool to "...store a data file and provide a download link"? If you don't have a recommendation, that's OK. I'm looking for guidance. Thanks again!
Anshika Varshney 15,620 Reputation points Microsoft External Staff Moderator

2026-01-09T18:38:46.5866667+00:00
Hi Michael Barnett,

Thank you for clarifying your requirement! If your goal is to store a data file and provide a secure download link within Microsoft Foundry, here’s the some approach:

Use Azure Blob Storage Integrated with Foundry

Blob Storage is ideal for storing files of any size and generating secure download links.

Steps:

Create a Storage Account in the https://portal.azure.com.

Add a Blob Container for your files.

Upload the Data File to the container.

Generate a Shared Access Signature (SAS) URL for secure download.

This SAS link can be returned by your Foundry workflow dynamically.

Integration with Foundry:

Register Blob Storage as a custom tool in Foundry.

Use Azure Storage SDK or REST API for programmatic access.

For low-code scenarios, leverage Power Platform connectors.

References

https://learn.microsoft.com/azure/storage/blobs/storage-quickstart-blobs-portal

https://learn.microsoft.com/azure/storage/common/storage-sas-overview

Blob Storage is the most scalable and secure option for production. If you prefer a simpler approach for testing, OneDrive or SharePoint can also work.

I hope this helps you get back on track! If you're still facing issues, could you share more details?

Thankyou!

Answer accepted by question author

Anshika Varshney 15,620 Microsoft External Staff Moderator

Hi Michael Barnett,

If you’re working with Microsoft Foundry, you need to try this approach is to leverage the Foundry Tools catalog and related services that make building and extending AI agents easier. Here are some practical options:

1. Use Foundry Tools Catalog

Foundry provides a central hub of tools (including Model Context Protocol – MCP servers) that you can integrate into your agents for tasks like:
- Data integration (connectors for SAP, Salesforce, HubSpot, etc.)
- Multimodal capabilities (speech, vision, document intelligence)
- Custom tool extensibility via APIs and Azure Logic Apps
You can also create a private tools catalog for organization-specific tools. Learn more about Foundry Tools

2. Foundry Agent Service

If your goal is orchestrating AI agents, Foundry Agent Service offers:
- Pre-built connectors to 1,400+ enterprise systems
- Secure orchestration and deployment of agents
- Integration with frameworks like LangChain or Microsoft Agent Framework Details on Agent Service

3. SDK & Developer Tools

For code-first development:
- Use the Foundry SDK (available in Python, Java, TypeScript, C#)
- Integrate with Visual Studio, GitHub, and Copilot Studio for streamlined workflows Quickstart guide

4. Integration Patterns

Foundry supports:
- Low-code connectors (Power Platform, Logic Apps)
- Direct REST API calls for full control
- API gateway mediation via Azure API Management for governance Integration guide

Start with the Foundry portal for quick prototyping and then scale using SDKs and orchestration services. This gives you flexibility to build declarative or code-first agents and integrate advanced tools as needed.

Please let me know if there are any remaining questions or additional details, I can help with, I’ll be glad to provide further clarification or guidance.

Thankyou!

Anshika Varshney 15,620 Reputation points Microsoft External Staff Moderator

2026-01-12T18:43:30.2733333+00:00

Hi Michael Barnett,

Thank you for sharing the additional details! It’s great to hear that you identified the root cause agent ignoring the URL and modifying blob names, which led to SAS token mismatches. Your observation about prompt adjustments improving robustness is spot-on, extensive tuning often makes agents more reliable in complex workflows.

I’ve converted my comment into an answer for better visibility. Please mark it as Accepted if this helped resolve your issue, so others facing similar problems can easily find the solution.

Thank you!

1 additional answer

Your answer

Michael Barnett 40 Reputation points

2026-01-08T17:30:17.0366667+00:00

We're using the New Foundry UI, but if there are configurations in the old UI, please provide details.
Michael Barnett 40 Reputation points

2026-01-09T15:24:56.8233333+00:00

Thanks Anshika Varshney; however, this response (generated by an LLM?) is non-specific. Can you please address the specific question about a tool to "...store a data file and provide a download link"? If you don't have a recommendation, that's OK. I'm looking for guidance. Thanks again!
Anshika Varshney 15,620 Reputation points Microsoft External Staff Moderator

2026-01-09T18:38:46.5866667+00:00

Hi Michael Barnett,

Thank you for clarifying your requirement! If your goal is to store a data file and provide a secure download link within Microsoft Foundry, here’s the some approach:

Use Azure Blob Storage Integrated with Foundry

Blob Storage is ideal for storing files of any size and generating secure download links.

Steps:

Create a Storage Account in the https://portal.azure.com.

Add a Blob Container for your files.

Upload the Data File to the container.

Generate a Shared Access Signature (SAS) URL for secure download.

This SAS link can be returned by your Foundry workflow dynamically.

Integration with Foundry:

Register Blob Storage as a custom tool in Foundry.

Use Azure Storage SDK or REST API for programmatic access.

For low-code scenarios, leverage Power Platform connectors.

References

https://learn.microsoft.com/azure/storage/blobs/storage-quickstart-blobs-portal

https://learn.microsoft.com/azure/storage/common/storage-sas-overview

Blob Storage is the most scalable and secure option for production. If you prefer a simpler approach for testing, OneDrive or SharePoint can also work.

I hope this helps you get back on track! If you're still facing issues, could you share more details?

Thankyou!
Anshika Varshney 15,620 Reputation points Microsoft External Staff Moderator

2026-01-12T18:43:30.2733333+00:00

Hi Michael Barnett,

Thank you for sharing the additional details! It’s great to hear that you identified the root cause agent ignoring the URL and modifying blob names, which led to SAS token mismatches. Your observation about prompt adjustments improving robustness is spot-on, extensive tuning often makes agents more reliable in complex workflows.

I’ve converted my comment into an answer for better visibility. Please mark it as Accepted if this helped resolve your issue, so others facing similar problems can easily find the solution.

Thank you!

Answer 1

Michael Barnett 40

Thanks Anshika Varshney, those are the technologies we're using in the tool we prepared and, as further information, the problem we experienced was due to the agent ignoring the URL returned by the tool. The agent was changing the timestamp portion of blob names (likely rounding or reformatting?), then using the SAS tokens with the wrong blob names, causing "Signature did not match" errors. More tweaks to the system prompts fixed the issue.

We are learning that robust agents result from extensive prompt adjustments.

0 comments