Connect an AI client to the OpenLM MCP Connector

Follow the instructions for your AI client to connect it to the OpenLM MCP Connector.

Before you begin

Confirm these prerequisites:

• An active OpenLM Platform account with access to a cloud-hosted tenant (US or EU).
• A paid AI subscription, such as Claude Pro, ChatGPT Plus, or equivalent. Free-tier plans don't support custom Model Context Protocol (MCP) connectors.
• The OpenLM MCP Connector URL for your region.

The following table lists the OpenLM MCP Connector URL for each environment:

Environment	OpenLM MCP Connector URL
Prod US	https://cloud-us.openlm.com/mcp
Prod EU	https://cloud-eu.openlm.com/mcp

Note

Replace the URL in each configuration that follows with the URL for your region.

Connect Cursor

This section applies to the Cursor desktop app.

1. Download and install Cursor from the official website.

2. Open Cursor.

3. Select Cursor Settings, then Tools and MCP.

4. Select Add a custom MCP server.

5. In the mcp.json file that opens, paste the following JSON configuration:

   {
     "mcpServers": {
       "openlm-reporting": {
         "url": "https://cloud-us.openlm.com/mcp"
       }
     }
   }

6. Sign in with your OpenLM Platform credentials.

7. Check that the OpenLM MCP Connector name and the total number of available tools appear.

8. Press Control+L to start a new chat and begin querying.

Connect Windsurf

This section applies to the Windsurf desktop app.

1. Download and install Windsurf from the official website.

2. Open Windsurf.

3. Select Windsurf Settings, then Open MCP Marketplace.

4. Select Add Custom MCP.

5. In the mcp_config.json file that opens, paste the following JSON configuration:

   {
     "mcpServers": {
       "openlm-reporting": {
         "command": "npx",
         "args": [
           "-y",
           "mcp-remote@latest",
           "https://cloud-us.openlm.com/mcp",
           "9877"
         ]
       }
     }
   }

6. Sign in with your OpenLM Platform credentials.

7. Check that the OpenLM MCP Connector name and tool count appear in the MCP Marketplace.

8. Open the chat panel and start querying.

Connect LibreChat

This section applies to the LibreChat web app.

1. Set up LibreChat from GitHub and open it in your browser.

2. In the side panel, select MCP Settings.

3. Select the + icon to add a new MCP server.

4. Enter a name for the OpenLM MCP Connector and the OpenLM MCP Connector URL for your environment.

5. Select Streamable HTTP as the transport type.

6. Select I trust the application.

7. Select Create.

   LibreChat redirects you to the browser for authentication.

8. Sign in with your OpenLM Platform credentials.

   After you sign in, the OpenLM MCP Connector is connected.

Connect Gemini CLI

This section applies to the Gemini CLI.

1. Install Gemini CLI globally by running this command:

   npm install -g @google/gemini-cli@latest

2. Run gemini in your terminal to initialize it for the first time.

3. Open the settings file at ~/.gemini/settings.json and paste the following configuration:

   {
     "mcpServers": {
       "openlm-reporting": {
         "command": "npx",
         "args": [
           "-y",
           "mcp-remote@latest",
           "https://cloud-us.openlm.com/mcp",
           "9877"
         ],
         "timeout": 30000
       }
     },
     "security": {
       "auth": {
         "selectedType": "oauth-personal"
       }
     }
   }

4. Save the file.

5. Run gemini again.

   The terminal shows openlm-reporting is connected, which confirms the connection.

6. Start querying tenant-specific data directly from the Gemini CLI terminal.

Connect Claude Code

This section applies to Claude Code in the CLI.

1. Open your terminal and run this command to add the OpenLM MCP Connector:

   claude mcp add --transport http openlm-reporting https://cloud-us.openlm.com/mcp

2. Authenticate when prompted.

   After you authenticate, the OpenLM MCP Connector is connected and ready to use.

3. Start querying tenant-specific reporting data directly from Claude Code.

Connect ChatGPT

This section applies to the ChatGPT web app.

Note

To access the Create Apps option, turn on Developer Mode in the ChatGPT settings.

1. Open ChatGPT in your browser and sign in to your account.

2. Select Settings, then Apps, then Create Apps.

3. Enter a name for the OpenLM MCP Connector and paste the OpenLM MCP Connector URL for your environment.

4. Select OAuth as the authorization type.

5. Select the confirmation check box.

6. Select Create.

   ChatGPT redirects you to the browser for authentication.

7. Sign in with your OpenLM Platform credentials.

   After you sign in, the OpenLM MCP Connector is connected.

8. Start a new chat, select the + icon, select More Options, and select the OpenLM MCP Connector to begin querying.

   The OpenLM MCP Connector appears beside the + icon.

Connect Claude.ai

This section applies to the Claude.ai web app.

1. Open claude.ai in your browser and sign in to your account.

2. Select your profile icon, then Settings, then Connectors.

3. Select Add custom connector.

4. Enter a name for the OpenLM MCP Connector and paste the OpenLM MCP Connector URL for your environment.

   Claude.ai redirects you to the browser for authentication.

5. Sign in with your OpenLM Platform credentials.

   After you sign in, the OpenLM MCP Connector is connected as a custom connector. You can use it directly in any Claude.ai conversation.

Tip

To query, select Connectors in the chat input area, select the OpenLM MCP Connector, and enter your question.

Connect Grok

This section applies to the Grok web app.

1. Open grok.com/connectors in your browser and sign in to your account.

   The Skills and Connectors page opens on the Connectors tab.

2. Select New Connector.

3. In the New Connector dialog, select Custom to add your own custom connector.

4. In the Custom Connector dialog, enter a name for the OpenLM MCP Connector and the OpenLM MCP Connector URL for your environment.

5. Select Add Connector.

6. On the OpenLM sign-in page, sign in with your OpenLM Platform credentials.

   After you sign in, the OpenLM MCP Connector appears under Installed on the Skills and Connectors page.

Tip

To query, turn on the connector in the chat. In the Grok chat window, select the + icon, select Connectors, and turn on the OpenLM MCP Connector.

Connect Microsoft Copilot Studio

This section applies to the Microsoft Copilot Studio web app. Microsoft Copilot connects through Copilot Studio, where you add the OpenLM MCP Connector as a tool of a custom agent.

Access Copilot Studio and create an agent

To create an agent for the OpenLM MCP Connector:

1. Open copilotstudio.microsoft.com and sign in with your Microsoft work account.
2. Check that the correct Environment for your organization is selected in the Environment selector.
3. Create a new agent, or open an existing one.
4. Enter a name for the agent, for example openlm-mcp.

Add the OpenLM MCP Connector as a tool

To add the OpenLM MCP Connector as a tool:

1. Open your agent and select the Tools tab in the top navigation.

2. Select + Add a tool, then select + New tool.

   This registers a new MCP server rather than picking a pre-built one.

3. Select Model Context Protocol as the tool type.

4. Enter the MCP server details:

   • Server name: a name for the connector, for example openlm-mcp.
   • Server description: a short description of what the server does.
   • Server URL: the OpenLM MCP Connector URL for your environment. The connector uses the Streamable HTTP transport.
   • Authentication: select OAuth 2.0, with the type Dynamic discovery.

5. Select Create.

   This registers the OpenLM MCP Connector and creates a connection.

6. On the OpenLM sign-in page, sign in with your OpenLM Platform credentials and authorize the connection.

Turn on and verify the tool

To turn on the tool and add it to your agent:

1. In the Tools list under Model Context Protocol, find the new tool.
2. Check that its Enabled toggle is set to On.
3. If you created the tool outside your agent, open the agent, select its Tools tab, and add the new tool to the agent.

Test the agent

To test the agent:

1. Select Test to open the Test your agent panel.
2. In the chat box, enter a prompt that uses the OpenLM tools, for example Show me license usage for the last month or List available OpenLM tools.
3. Confirm that the agent calls the OpenLM MCP Connector and returns the results in the chat.

Note

The MCP tool must be Enabled and the connection authenticated. Otherwise, the agent can't call the OpenLM tools.

What else you can do in Copilot Studio

After you connect the OpenLM MCP Connector, you can use the following Copilot Studio features:

Feature	What you can do
Test	Live-test the agent's real responses against the OpenLM tools in the built-in chat panel before publishing.
Evaluation	Evaluate the agent's answer quality and accuracy across multiple test cases.
Agents	Build multiple specialized agents, each with its own instructions and tools.
Topics	Define conversational flows and triggers that guide the agent's behavior for specific user intents.
Flows (Power Automate)	Build workflows that combine the OpenLM tools with other actions, such as sending email or triggering automations.
Knowledge	Add knowledge sources, such as documents or websites, so the agent can answer beyond the OpenLM tool data.
Channels	Publish the agent to Microsoft Teams, a website, Copilot chat, or other channels.
Monitor	Track usage, conversations, and performance analytics of the agent.
Publish	Push the agent live so your team or organization can use it.

Connect GitHub Copilot (VS Code)

This section applies to the GitHub Copilot desktop app in VS Code. GitHub Copilot reads MCP servers from a configuration file in the project.

1. In VS Code, open any project folder.

2. In the project root, create a folder named .vscode if it doesn't already exist.

3. In .vscode, create a file named mcp.json.

4. Paste the following configuration, entering a name for the server and the OpenLM MCP Connector URL for your environment:

   {
     "servers": {
       "openlm-mcp-copilot": {
         "type": "http",
         "url": "https://cloud-us.openlm.com/mcp"
       }
     }
   }

5. Open the Command Palette (Control+Shift+P).

6. Select MCP: List Servers.

7. Select your server, for example openlm-mcp-copilot, and select Start Server.

8. On the OpenLM sign-in page, sign in with your OpenLM Platform credentials.

   After you sign in, the server connects and the tools are discovered.

9. Open GitHub Copilot Chat.

10. Set the mode drop-down list at the bottom of the chat box to Agent.

    Tools work only in Agent mode.

Tip

In the Copilot chat box, select the tools icon and check that the tools of the OpenLM MCP Connector are toggled on.

Connect Cline

This section applies to the Cline desktop app. Cline is a VS Code extension that reads MCP servers from its own settings file.

1. In VS Code, install the Cline extension from the Extensions marketplace.

2. Open Cline and select the MCP Servers icon in the top toolbar.

3. Select the Configure tab, then Configure MCP Servers.

   The cline_mcp_settings.json file opens.

4. Paste the following configuration:

   {
     "mcpServers": {
       "openlm-reporting": {
         "type": "streamableHttp",
         "url": "https://cloud-us.openlm.com/mcp"
       }
     }
   }

5. Save the file.

   The OpenLM sign-in page opens.

6. If the page doesn't open automatically, select Retry Connection.

7. Sign in with your OpenLM Platform credentials.

   After you sign in, the OpenLM MCP Connector and its tools appear in the MCP Servers list.

Tip

You can also add the connector without editing JSON. On the MCP Servers panel, select the Remote Servers tab, enter a server name and the OpenLM MCP Connector URL for your environment, select the Streamable HTTP transport type, and select Add Server.

Connect Codex

This section applies to the Codex desktop app.

1. Download Codex from the official website and install it.

2. Complete the initial setup and sign in to your account.

3. Select Settings, then MCP servers, then Add server.

4. Enter the server details:

   • Name: a name for the connector, for example openlm-reporting.
   • Transport: select Streamable HTTP.
   • URL: the OpenLM MCP Connector URL for your environment.

5. Save the configuration.

6. If Codex prompts you, restart it.

7. Select Authenticate.

   The OpenLM sign-in page opens.

8. Sign in with your OpenLM Platform credentials.

   After you authenticate, the OpenLM MCP Connector is connected and its tools are available.

Note

The same steps apply in the Codex IDE extension, which uses the same Settings, then MCP servers panel.

Connect Google Antigravity

This section applies to the Google Antigravity desktop app. Antigravity reads MCP servers from an mcp_config.json file.

1. Download and install Antigravity from the official website.

2. In the Antigravity IDE, select the … menu in the agent side panel, then MCP Servers.

3. Select the option to edit the configuration file directly.

4. In mcp_config.json, paste the following configuration:

   {
     "mcpServers": {
       "openlm-reporting": {
         "command": "npx",
         "args": [
           "-y",
           "mcp-remote@latest",
           "https://cloud-us.openlm.com/mcp",
           "9880"
         ]
       }
     }
   }

5. Save the file.

   The OpenLM sign-in page opens.

6. Sign in with your OpenLM Platform credentials.

   After you sign in, the OpenLM MCP Connector and its tools appear in the MCP servers list.

Note

The number after the URL (9880) is the local port that mcp-remote uses for the OAuth callback. If you run more than one MCP server through mcp-remote, give each one a different port so that they don't collide.

Tip

For best performance, Antigravity recommends keeping about 50 MCP tools enabled across all your MCP servers. The OpenLM MCP Connector exposes 19 tools, so you don't normally need to trim it. If you connect several MCP servers at once and need to reduce the count, add a disabledTools array to the OpenLM entry that lists the tool names to withhold from the model. For the tool names, see Tools and capabilities reference.