Claude (desktop app and browser), ChatGPT, and Microsoft Copilot are documented for connecting to GLM/SLM through MCP today. Microsoft Copilot setup has two parts: an IT administrator configures a Copilot Studio agent, and then each end user connects to that agent with their own credentials.

Reading attachments is available only through Claude Code and ChatGPT Codex. For details, see Can AI Tools Read Attachments in GLM or SLM?

Confirm the following before connecting an AI chat tool to a GLM or SLM site through MCP:

• Access to Claude (desktop app installed and signed in, or the browser version at claude.ai), ChatGPT (browser version at chatgpt.com, or the desktop app), or Microsoft Copilot (Microsoft 365 Copilot at m365.cloud.microsoft/chat, or Copilot in Teams), depending on the AI chat tool being connected.
• Admin credentials for the GLM or SLM site.
• The correct MCP URL for the environment:
  • Sandbox site: https://mcp.data.sandbox.grantinterface.com/mcp
  • Live site: https://mcp.data.grantinterface.com/mcp
• For Microsoft Copilot only:
  • A Microsoft 365 Copilot license for each end user who will use the agent.
  • A Copilot Studio User License and the Copilot Studio Authors role in the Power Platform environment for the IT administrator who builds the agent.

These steps work the same way in the Claude desktop app and in the browser at claude.ai.

To connect a GLM or SLM site to Claude using MCP:

1. Open the Claude desktop app, or go to claude.ai in a browser, and sign in.
2. Click Customize in the left side menu.
3. Click Connectors in the sidebar, and then click the plus icon at the top of the Connectors column.
4. In the Add custom connector dialog, enter GLM MCP in the Name field and the MCP URL for the environment in the URL field.
5. Click Add.
6. Click GLM MCP in the Not Connected list, and then click Connect.
7. On the Foundant login page that opens, enter the admin email and password for the GLM or SLM site.
8. Close the confirmation tab, or click Open Desktop App, to return to Claude.

Ask Claude About GLM or SLM Data

Once connected, start a new chat in Claude and ask a question about GLM or SLM data in plain language, examples are included below:

• List the top ten most recent grant or scholarship requests.
• Show applications submitted in the last 30 days.
• Summarize the awards made in the current fiscal year.

Custom MCP connectors in ChatGPT sit behind a Developer mode toggle and require a brief security acknowledgment. Setup takes about the same amount of time as Claude, roughly 3 to 5 minutes.

These steps work the same way in the ChatGPT browser version at chatgpt.com and in the desktop app.

To connect a GLM or SLM site to ChatGPT using MCP:

1. Open chatgpt.com in a browser, or launch the ChatGPT desktop app, and sign in.
2. Click the account name at the bottom-left of the sidebar, and then click Settings.
3. Click Apps in the left sidebar of the Settings dialog.
4. Click Advanced settings on the Apps page.
5. Toggle Developer mode on. Leave Enforce CSP in developer mode unchecked, which is the default setting. A Create app button appears at the top-right of the page once Developer mode is on.
6. Click Back to return to the main Apps page, and then click Create.
7. Add the applicable information to the New App dialog. Asterisks indicate a required field. Advanced OAuth settings do not need to be changed; the defaults are correct.
   • Name - GLM MCP
   • Description (optional) - Access my GLM data
   • MCP Server URL - The MCP URL for the environment being connected
   • Authentication - Select OAuth from the drop-down menu
8. Check the I understand and want to continue box, and then click Create.
9. Enter the admin email and password for the GLM or SLM site on the Foundant login page that opens.
10. Confirm that ChatGPT returns with a "GLM MCP is now connected" banner, and that GLM MCP appears under Enabled apps.

Ask ChatGPT About GLM or SLM Data

Once connected, start a new chat in ChatGPT and ask a question about GLM or SLM data in plain language, examples are included below:

• List the top ten most recent grant or scholarship requests.
• Show applications submitted in the last 30 days.
• Summarize the awards made in the current fiscal year.

The first time ChatGPT uses the connector for a particular type of data, it displays a permission dialog describing the data being retrieved, for example, "Query GLM request records?" Review the details, and then click Query Data to allow it, or Deny to cancel. This per-action consent step is expected and keeps visibility into what data is shared.

Connecting Microsoft Copilot to a GLM or SLM site through MCP has two parts. A Microsoft IT administrator creates and publishes a Copilot Studio agent, and then each end user connects that agent to GLM or SLM data with their own credentials. Agent configuration is a one-time task completed in Copilot Studio, not in GLM or SLM.

Agent Configuration in Copilot Studio

To configure the agent, a Microsoft IT administrator creates, tests, and publishes a Copilot Studio agent in the organization's Microsoft 365 Copilot environment. After the agent is published, each end user connects to it with their own GLM or SLM credentials.

Configure the Agent

To create the Copilot Studio agent for GLM or SLM, sign in to Copilot Studio and enter the agent details:

1. Navigate to copilotstudio.microsoft.com and sign in with the Microsoft 365 admin account.
2. In the left-side menu, click Agents.
3. Click New agent.
4. Enter the following agent details:
   • Name - GLM Assistant
   • Instructions - You are a grants management assistant connected to GLM or SLM via live data tools. When a user asks about grants, scholarships, applications, requests, organizations, or awards, use the available tools to retrieve real-time information. Only report what the tools return. If a query is ambiguous, ask a clarifying question. Do not fabricate amounts, dates, or organization names.

Add the MCP Tool and Create the Initial Connection

To add GLM or SLM data to the Copilot Studio agent, add the MCP tool in the agent editor and create the initial connection:

1. In the agent editor, click Tools.
2. Click Model Context Protocol (MCP).
3. Click Add, and then click Model Context Protocol (MCP).
4. Enter the following MCP server details, and then click Add:
   • Server name - GLM MCP
   • Server description - GLM and SLM grants and scholarship management tools. Provides read access to grant and scholarship requests, organizations, and awards.
   • Server URL - The MCP URL for the environment being connected
     • Sandbox site: https://mcp.data.sandbox.grantinterface.com/mcp
     • Live site: https://mcp.data.grantinterface.com/mcp
   • Authentication - OAuth 2.0
   • Configuration type - Dynamic (with discovery)
5. Once the MCP server is successfully created, click Next.
6. Continue to the Add tool screen, and then click Create new connection for the GLM MCP server.
7. In the sign-in window, enter the admin email and password for the GLM or SLM site to verify that the connection works.
   • This is an admin verification step and does not affect how end users connect.
8. After a successful sign-in, click Add to agent to finish adding the GLM MCP server to the agent.

Validate the MCP Connection

To validate the MCP connection, run a test request in the Copilot Studio preview experience and verify the results:

1. Run a test request in the preview experience, such as "List the top ten most recent grant or scholarship requests."
2. Verify the following:
   • The GLM MCP tool is invoked.
   • Authentication succeeds.
   • The MCP request returns data.
   • The returned data matches the permissions of the signed-in GLM or SLM user.
   • No unexpected authorization or connection errors occur.

Publish and Share the Agent

To publish the GLM Assistant agent and share it with end users, add the Teams and Microsoft 365 Copilot channel, and then share the agent with the users or security groups that need access:

1. In the agent editor, click Publish in the upper navigation bar.
2. Click Channels, select Teams and Microsoft 365 Copilot, and then click Add channel.
3. Click the three dots icon, and then click Share.
4. Add the users or security groups that should have access to the agent and assign them the Viewer role.
5. Click Update.

End User Setup

After an IT administrator publishes and shares the Copilot Studio agent, each end user opens the GLM Assistant agent in Microsoft Copilot and connects it to GLM or SLM data. The first time the agent is used, Copilot prompts for the GLM or SLM sign-in. This connects the agent to the site's data using that user's own credentials.

Locate the GLM Assistant in Microsoft Copilot

To find the GLM Assistant agent in Microsoft Copilot:

1. Open Microsoft 365 Copilot by navigating to m365.cloud.microsoft/chat or opening Copilot in Teams.
2. In the left-side menu under Agents, click GLM Assistant.
   • If this option does not appear, click More agents to browse the agents available in the organization.

Connect Your GLM Account

To connect the GLM Assistant agent to a GLM or SLM account, sign in through the Connection Manager the first time the agent is used:

1. After opening the agent, click the Open Connection Manager link in the Connection Manager window.
2. Click Add new connection.
3. On the Foundant login page, enter the admin email and password for the GLM or SLM site, and then click Sign in.
4. Once signed in, click Submit. The Connection Manager confirms that the connection is active.
5. Return to the Copilot chat window, and click Retry to resubmit the original query.

Use the GLM Assistant

Once connected, ask the GLM Assistant a question about GLM or SLM data in plain language. The agent returns only the information that the signed-in user's permissions allow. Sample prompts include:

• List the top ten most recent grant or scholarship requests.
• Show applications submitted in the last 30 days.
• Summarize the awards made in the current fiscal year.
• How many requests were submitted in the last 7 days?

Is MCP Read-Only?

MCP is read only in its current state, with access to the same data available in Foundant Reporting. Accessing documents through MCP may require IT support on the organization's side.

Can AI Tools Read Attachments in GLM or SLM?

Reading GLM or SLM attachments through MCP requires Claude Code or ChatGPT Codex.

Is My Data Secure When Connected via MCP?

Each connection is scoped to the credentials the administrator sets up during connection, so only that site's own data is exposed. Only users with the full Administrator role can connect to MCP, and existing portal permissions are honored through the connection.

Can SLM Clients Use the Same Connector as GLM?

Yes, SLM clients use the same MCP connector as GLM. Each product requires its own authentication.

Troubleshoot a Claude MCP Connection

• Sign-in fails or loops back to the login page - Confirm the admin credentials match the correct GLM or SLM site, and verify the URL matches the environment (Sandbox or Live).
• Browser does not open after clicking Connect - Confirm the default browser is set and pop-ups are not blocked for claude.ai.
• "Not connected" persists after signing in - Return to the Connectors panel and click Connect again. If the issue continues, remove the connector and add it again using the correct URL.
• Wrong environment connected - Remove the existing GLM MCP connector, and add a new one using the correct sandbox or live URL.

Troubleshoot a ChatGPT MCP Connection

• "Create" button does not appear - Confirm Developer mode is toggled on under Apps, Advanced settings, and that the main Apps page has been returned to after enabling it.
• Sign-in fails or loops back to the login page - Confirm the admin credentials match the correct GLM or SLM site, and verify the URL matches the environment (sandbox or live).
• Browser does not redirect to the GLM login page - Confirm pop-ups are not blocked for chatgpt.com. If the redirect fails, click Create again from the New App dialog.
• "Not connected" persists after authentication - Return to the Apps page and check the Drafts and Enabled apps sections. If GLM MCP appears in Drafts only, click into it to complete setup. As a last resort, delete the connector and add it again.
• Wrong environment connected - Delete the existing GLM MCP connector, and add a new one using the correct sandbox or live URL.
• Permission dialog appears every time - This is expected for new types of data requests. ChatGPT asks for explicit consent before each new kind of operation to keep the administrator in control of what is accessed.

Get Accurate Results

If a result from MCP looks unexpected, compare it against the same information in Foundant Reporting. It can also help to tell the AI tool which data source to use, since it may need direction to use the approval data set for grant amounts instead of payment data, for example.