Plug & play, production-ready. Build your agents, workflows, and logic in Azure AI Foundry—then plug them into this app. No custom front end to host or maintain: add the web part to a SharePoint page or a Teams tab, configure your agents, and you’re live. One tool, one place, for all your Foundry agents.

Stay inside your perimeter. Unlike generic chat UIs or self-hosted solutions (e.g. LibreChat), your users never leave your intranet. They work in the same SharePoint sites and Teams they already use. No redirect to external portals, no new URLs to remember—AI is just another part of the page or the tab.

Data sovereignty matters. Conversations and calls go through your Microsoft 365 and Azure tenant. Your data stays in your ecosystem, under your governance and compliance rules. For organizations that care deeply about where data lives and who processes it, this is a decisive advantage: enterprise AI without sending traffic outside your controlled environment.

A single, scalable front door. One app to expose many agents (HR, IT, support, knowledge, etc.), with audience and permissions managed in one place. The possibilities are wide open: whatever you design in Azure AI Foundry can be surfaced here, in the tools your people already use every day.

End users: Discover and chat with AI agents from SharePoint pages or Teams.
Page owners / authors: Add the web part and choose which agents appear (if they have edit rights).
Administrators: Deploy the app, approve required API permissions, and configure Azure AI projects and roles so the app works correctly.
Key capabilities
Works in SharePoint and Teams
The same app runs as:

A SharePoint web part on classic or modern pages, and in SharePoint full-page apps.
A Microsoft Teams personal app or Teams tab in a channel.
One configuration works across both; you can reuse the same agents on SharePoint and in Teams.

Dynamic theme (SharePoint and Teams)
The app automatically follows the look and feel of the host:

SharePoint: Uses the theme of the SharePoint page (colors, fonts) so the chat blends with your site.
Teams: Uses the Teams theme (default, dark, or high contrast) and updates if the user changes it.
No extra setup is required for themes.

Multiple agents in one place
Admins and page authors configure several agents in a single web part using the built-in property pane. For each agent you can set:

Display name and icon (Fluent icon or custom image).
Agent ID and Endpoint URL (from your Azure AI Foundry project).
Suggestion prompts (shortcuts shown when starting a new chat).
Conversation history: show only conversations started in this web part, or all conversations for that agent.
Attachments: allow or disallow file attachments in chat.
Audience (see below).
So one web part can expose many agents (e.g. HR, IT, Sales) with different settings per agent.

Audience per agent
For each agent you can restrict who can see and use it:

Everyone: leave audience empty → all users who can open the page or tab see the agent.
Specific people: allow only certain users, Entra (Azure AD) groups, or SharePoint site groups.
Only users who belong to at least one of the selected audiences see that agent in the list and can open its chat. Others don’t see it at all. This is configured per agent in the web part properties.

For end users: seeing the app vs. chatting
Seeing the app does not automatically mean you can chat with an agent.

To see the web part you only need access to the SharePoint page or the Teams tab (normal site/team permissions).
To actually send messages and get replies you need the “Azure AI User” role (or equivalent) on the Azure AI project that backs that agent.
If you don’t have that role, you may see the agent in the list but get errors when you try to chat. Your admin must assign you the Azure AI User role on the relevant Azure AI project (in the Azure portal). There is no way for the app itself to grant that permission.

For administrators: deployment and API approval
Approving API permissions is required
After the app is deployed (e.g. to the tenant app catalog), a tenant admin must approve the API permissions requested by the app. Until they are approved, the app will not work correctly and users can get errors when opening or using it.

Azure Machine Learning Services — scope: user_impersonation (used to call the Azure AI Foundry agent APIs on behalf of the signed-in user).
Microsoft Graph — scope: User.Read (used for the signed-in user’s profile and for audience checks).
Approval is done in the SharePoint admin center (or equivalent) under the deployment / API access / permission requests for this solution. The admin must accept these permissions for the app to function.

Prerequisite: create an Azure AI Foundry project first
The “Azure Machine Learning Services” permission that the app needs is tied to the Azure AI / Machine Learning ecosystem. For the approval flow to succeed:

At least one Azure AI Foundry project must already exist in your tenant (or the subscription you use for SharePoint/Teams).
Creating that project ensures the “Azure Machine Learning Services” service principal is available in the tenant for consent.
Then the admin can approve the app’s request for Azure Machine Learning Services (user_impersonation).
If no Azure AI Foundry project has ever been created, the “Azure Machine Learning Services” permission may not appear as valid during approval. In that case, the admin may see:

“The requested permission isn’t valid. Reject this request and contact the developer to fix the problem and redeploy the solution.”
What to do: Create at least one Azure AI Foundry project (in the Azure portal) so that the Azure Machine Learning Services integration is provisioned. After that, run the app deployment again and have the admin approve the API request. Do not approve blindly; ensure the permission shows as valid (e.g. “Azure Machine Learning Services” with user_impersonation) before accepting.

Create at least one Azure AI Foundry project so “Azure Machine Learning Services” is available.
Deploy the Foundry Agents app (e.g. upload the package to the app catalog and deploy).
Approve the app’s API permissions (Azure Machine Learning Services + Microsoft Graph) in the admin center.
Assign the Azure AI User role on the relevant Azure AI project(s) to the users or groups who should be able to chat.
Configuration (admins and page authors)
Where: When editing the page (SharePoint) or the tab (Teams), select the Foundry Agents web part and open its properties (pane on the side).
Agents: Add, edit, reorder, or remove agents. Each agent has:
Display name, Agent ID, Endpoint URL (from Azure AI Foundry).
Icon (Fluent icon or custom image URL).
Suggestion prompts, conversation history scope, attachments on/off.
Audience: optional list of users/Entra groups/SharePoint groups; empty = everyone.
Appearance: Optional minimum height (e.g. min(600px, 90vh)).
No secrets are stored in the web part; authentication uses the current user’s Microsoft 365 identity and the approved API permissions.

User identity: The app uses the signed-in user’s Microsoft 365 account to call Azure AI Foundry. No API keys or passwords are entered in the app.
Who can chat: Governed by Azure AI project roles (e.g. Azure AI User). Grant these in the Azure portal for the right project(s).
Who sees which agent: Governed by audience in the web part (users/groups) and by normal SharePoint/Teams access to the page or tab.
What you need to use it
Role Needs
End user Access to the page or Teams tab; Azure AI User (or equivalent) on the Azure AI project for each agent they use.
Page/tab author Edit rights on the page or tab; same as end user if they want to chat.
Admin Ability to deploy the app, approve API permissions, and create/configure Azure AI projects and assign Azure AI User role.
Tech stack
SharePoint Framework (SPFx)
React, Fluent UI
Microsoft Foundry Agents / Azure AI APIs
Microsoft 365 identity and permissions
Status
In development, feedback and contributions are welcome.

We will keep developing and improving this tool in step with Azure AI Foundry, which is evolving very quickly. The goal is to stay aligned with new capabilities as they become available—so you keep one front door that grows with the platform.

Possible directions for future releases include (non-exhaustive):

Streaming: Direct streaming of agent responses for a more responsive chat experience.
Voice: Use voice input instead of (or in addition to) typing.
Browser automation: Integration with agent capabilities that drive or automate actions in the browser.
Richer agent features: As Azure AI Foundry adds new tools and APIs, we will evaluate exposing them in the app.
If you have ideas or needs, opening an issue or contributing is the best way to shape this roadmap.

Contributing
Contributions, issues, and feature requests are welcome to improve this app for SharePoint and Teams.