Hello, Our team recently adopted AI Foundry as a tool for semi-technical people to collaborate on AI efforts. One such aspect is the usage of promptflow for building LLM-powered workflows. The tool is being well received apart from one crucial aspect: collaboration. I can't seem to figure out how to configure the AI Foundry project to allow multiple people to collaborate on a single Promptflow flow. This is the view from the creator of the flow: And this. is the view of a person with the "Azure AI developer" role in AI Foundry: On top it says that there is a need to clone it to edit it. However, in the Azure documentation, AI Foundry promptflow is heavily framed as a collaborative tool (eg: https://learn.microsoft.com/en-us/azure/ai-foundry/how-to/prompt-flow#enterprise-readiness ). Could you share if there is a way or workaround so that multiple people can edit the same flow? Or if not, could you share if this is on the roadmap? Kind regards, Thomas

Hi Thomas Dehaene To enable collaboration on a Promptflow flow in Azure AI Foundry. You can add respective users on IAM of AI studio project. But multiple users cannot edit the same flow simultaneously, which can be a limitation for teams looking for real-time collaboration. Here are some practices: Set up a centralized code repository to manage and track changes effectively. Use local environments for individual testing before merging changes into the main flow. Implement version control practices to manage different iterations of the flow. Kindly refer below link: rbac-ai-foundry Thank You.

How to allow collaboration on a promptflow flow in Azure AI Foundry

Thomas Dehaene 10

Hello,

Our team recently adopted AI Foundry as a tool for semi-technical people to collaborate on AI efforts.

One such aspect is the usage of promptflow for building LLM-powered workflows. The tool is being well received apart from one crucial aspect: collaboration.

I can't seem to figure out how to configure the AI Foundry project to allow multiple people to collaborate on a single Promptflow flow.

This is the view from the creator of the flow:

User's image

And this. is the view of a person with the "Azure AI developer" role in AI Foundry:

User's image

On top it says that there is a need to clone it to edit it.

However, in the Azure documentation, AI Foundry promptflow is heavily framed as a collaborative tool (eg: https://learn.microsoft.com/en-us/azure/ai-foundry/how-to/prompt-flow#enterprise-readiness).

Could you share if there is a way or workaround so that multiple people can edit the same flow?

Or if not, could you share if this is on the roadmap?

Kind regards,

Thomas

Thomas Dehaene 10 Reputation points

2025-03-04T12:51:36.13+00:00
Hi Saideep,

Thanks for the swift reply! Though I'm afraid it doesn't solve my case

I understand that realtime collaboration is out of the question. But non-realtime collaboration on the same promptflow also seems impossible right? As in:

person A goes to AI foundry and opens flow X

they start up a compute session, get to work, do edits, runs and saves

person A exits AI foundry

person B can open flow X

person B starts up a compute session, get to work, do edits and saves

That is not possible right? Or is it?

Also appreciate your other pointers, but as I see it, it is currently impossible to connect AI foundry promptflow to a version control system right?
Thomas Dehaene 10 Reputation points

2025-03-04T12:54:30.9866667+00:00

thanks in advance
Saideep Anchuri 3,850 Reputation points Microsoft External Staff

2025-03-04T13:16:22.69+00:00
Hi Thomas Dehaene

person A goes to AI foundry and opens flow X

they start up a compute session, get to work, do edits, runs and saves

person A exits AI foundry

person B can open flow X

person B starts up a compute session, get to work, do edits and saves - This is possible.

Thank You.
Thomas Dehaene 10 Reputation points

2025-03-04T13:21:49.42+00:00

Hi @Saideep Anchuri ,

Thanks for the reply.

So in the described scenario, you say that it is possible, but when I try it out, person B get the visual like in my second screenshot, where he can only "clone" the flow, there is no button to start a compute session.

Is this an RBAC issue? If so, can you point me to the needed IAM/RBAC role to have those buttons visible?
Saideep Anchuri 3,850 Reputation points Microsoft External Staff

2025-03-04T14:09:23.6966667+00:00

Hi Thomas Dehaene

Ideally It should be Azure AI Developer to start computes, use AI search and create deployment (Except for creating new hubs).

Reference - Role-based access control in Azure AI Foundry portal - Azure AI Foundry | Microsoft Learn

Thank You.
Thomas Dehaene 10 Reputation points

2025-03-04T14:22:57.3366667+00:00

Hi @Saideep Anchuri , that is no help either I'm afraid.

In my setup, both person A and person B have the Azure AI Developer role.

However, person B can only start computes and start runs in flows that he created.

In flows that person A created, person B can only see the "clone" button.
JAYA SHANKAR G S 900 Reputation points Microsoft External Staff

2025-03-05T04:52:15.13+00:00

Hi @Thomas Dehaene ,

Dose person B can be able to create new prompt flow like how person A did in the initial?
If he is able to do then you have correct roles, and regarding editing exiting prompt flow will check and update.

Thank you
JAYA SHANKAR G S 900 Reputation points Microsoft External Staff

2025-03-06T03:39:07.99+00:00

Hi @Thomas Dehaene ,

I haven't heard from you, did you try the above suggestion?

Let me know if you any query.

Thank you
JAYA SHANKAR G S 900 Reputation points Microsoft External Staff

2025-03-06T08:10:21.9266667+00:00

Hi @Thomas Dehaene ,

It seems like in portal the owned the user can be able to make changes.

Try using it in VS code extension for better version control and collaboration.

It's also mentioned in above github repository

Collaborate with your team by leveraging the cloud version of Prompt flow in Azure AI.

Which is pointing to prompt flow in machine learning workspace, so check that also.

Let me know if any further queries regarding this.

Thank you
Thomas Dehaene 10 Reputation points

2025-03-07T07:25:53.29+00:00

Hi @JAYA SHANKAR G S , that's not a solution either I'm afraid.

Using ML Workspaces, and accessing promptflow from there, person B still cannot edit a promptlfow that has been created by person A (see screenshot from person B), person B still can only clone it.

VS Code extension is not an option for us, as we want to have a business-friendly interface to maintain and evolve LLM-based applications.
JAYA SHANKAR G S 900 Reputation points Microsoft External Staff

2025-03-07T08:03:14.4633333+00:00

Hi @Thomas Dehaene ,

Please check what roles you have given to Users in Management Center and also try giving contributor role.

Check this and follow the steps. Please add the snippet of the like below.

Thank you
Thomas Dehaene 10 Reputation points

2025-03-07T12:22:14.2366667+00:00
Hi @JAYA SHANKAR G S : again this does not change anything.

If I give person B (Thomas Dehaene)

Contributor rights on the hub and

Azure AI Developer rights on the project

Still, person B cannot edit a flow created by person A:

Also for that matter, any combination of rights for person B (also owner on the project and owner on the hub) does not result in a combination where person B can start a compute session, do edits and saves in a flow from person A.

If you indeed have a working combo on your end, can you share the RBAC setup (both on Azure and inside the Foundry) that works please?
JAYA SHANKAR G S 900 Reputation points Microsoft External Staff

2025-03-07T14:43:37.3533333+00:00

Hi @Thomas Dehaene , we are checking with internal team, will update once we get the details.

Thank you

1 answer

Saideep Anchuri 3,850 Reputation points Microsoft External Staff

2025-03-04T09:41:19.9933333+00:00
Hi Thomas Dehaene

To enable collaboration on a Promptflow flow in Azure AI Foundry. You can add respective users on IAM of AI studio project. But multiple users cannot edit the same flow simultaneously, which can be a limitation for teams looking for real-time collaboration.

Here are some practices:

Set up a centralized code repository to manage and track changes effectively.

Use local environments for individual testing before merging changes into the main flow.

Implement version control practices to manage different iterations of the flow.

Kindly refer below link: rbac-ai-foundry

Thank You.
Please sign in to rate this answer.

0 comments No comments
Sign in to comment

Use comments to ask for clarification, additional information, or improvements to the question.

Share via

How to allow collaboration on a promptflow flow in Azure AI Foundry

1 answer

Your answer