Skip to main content

Configuring MCP for Luminary Cloud

Learn how to set up and use Luminary's MCP Server

Updated this week

What is MCP?

MCP (Model Context Protocol) is an open protocol that Anthropic introduced to standardize how applications provide context to large language models (LLMs). Think of it like USB-C for AI—just as USB-C gives you a standard way to connect devices to peripherals, MCP provides a standard way to connect AI models with various data sources and tools.

What is the Luminary Cloud MCP Server?

The Luminary Cloud MCP Server supercharges AI agents in your favorite IDEs like Visual Studio Code and Cursor. It gives them insider knowledge of the Luminary Cloud Platform and SDK, plus access to key platform API resources.

This integration dramatically improves your AI agents' ability to write, refactor, and debug SDK scripts. The result? You'll build simulation pipelines with the Luminary Cloud SDK much faster.

Our MCP Server uses the latest Streamable HTTP transport protocol and works seamlessly with popular IDEs (Visual Studio Code, Cursor) and AI agents (Claude Desktop, Claude Code, Gemini CLI).

How to Connect to the Luminary Cloud MCP Server

Visual Studio Code

Option 1: One-Click Install

The easiest way to get started:

  1. Open Visual Studio Code

  2. Click the install link:

  3. Click "Install" when VSCode shows the MCP Server installation window

  4. Open Copilot by selecting the Copilot icon near the search bar

  5. Don't see the Copilot icon? Make sure you've installed the "Copilot" extension first.

  6. Select "Agent" mode from the options

  7. Ask your first question about Luminary Cloud and its Python SDK. The authorization process will start automatically—just log in with your Luminary Cloud account.

  8. Verify everything works by asking the Agent if it can see the Luminary Cloud MCP. If not, check the Troubleshooting section below.

Option 2: Manual Configuration

If you prefer to set things up manually:

  1. Open the Command Palette and run MCP: Add Server

  2. Select "HTTP" as your server type

  3. Enter the server URL: https://app.luminarycloud.com/mcp

  4. Add a descriptive name: Luminary Cloud

  5. Choose your installation scope:

    • Global: Available in all workspaces, runs locally

    • Remote: Available on remote machine, runs on SSH

    • Workspace: Available in current workspace only

  6. Confirm the configuration in your mcp.json file:

    {
    "servers": {
    "Luminary Cloud": {
    "url": "https://app.luminarycloud.com/mcp",
    "type": "http"
    }
    }
    }​

  7. Allow authentication when the permission window appears, then complete the browser authorization process

  8. Open Copilot (the icon near the search bar) and make sure "Agent" mode is selected

  9. Complete authorization by clicking the circle icon and logging in with your Luminary Cloud account

Cursor

First, make sure you have Cursor downloaded from cursor.com. You might need a monthly subscription to enable the Cursor Agent.

Option 1: One-Click Install

  1. Open Cursor

  2. Click the install link or copy it into your browser:

  3. Click "Install" when Cursor switches to the "Tools & Integrations" page with the pre-filled configuration

  4. Complete authentication by clicking "Needs login" and following the browser authorization steps

  5. Verify the connection by checking that the yellow dot turns green, showing "5 tools enabled"

  6. If the dot stays yellow, try reloading the window from the Command Palette and repeat the authentication step.

Option 2: Manual Configuration

Note: Cursor's interface changes between versions, so these steps might vary slightly, but the core configuration remains the same.

  1. Open Cursor Settings:

    • Mac: Cursor → Settings → Cursor Settings

    • Windows: File → Preferences → Cursor Settings

    • Or: Click the gear icon in the top right corner

  2. Select "Tools & Integrations" from the left panel

  3. Click "New MCP Server" at the bottom of the page

  4. Add the configuration by copying this into the JSON file that opens:

    {
    "mcpServers": {
    "Luminary Cloud": {
    "url": "https://app.luminarycloud.com/mcp"
    }
    }
    }

  5. Save and close the JSON file

  6. Return to "Tools & Integrations" to see the Luminary Cloud server in your list

  7. Complete authentication by clicking "Needs login" and following the browser steps

  8. Confirm success when you see the green dot showing "5 tools enabled"

How to Use the MCP Server

Once you're connected, using the MCP Server is straightforward:

  1. Open your chatbot and make sure "Agent" mode is selected

  2. Ask questions like:

    • "Please list my Luminary Cloud projects"

    • "How do I create a project using the Luminary Cloud SDK?"

  3. Grant permission when the agent requests to use Luminary Cloud tools (this only happens the first time)

Pro Tips

Remind your agent to use MCP tools: Sometimes agents might write code from scratch instead of calling our MCP tools, which can lead to outdated or incorrect code. If this happens, try adding this reminder to your prompt:

"You're encouraged to call the Luminary Cloud MCP tools before writing or editing any Python code using the Luminary Cloud SDK."

Use custom instructions: Both VSCode and Cursor support custom instructions where you can add this reminder permanently:

Troubleshooting

Visual Studio Code Issues

The Copilot agent isn't calling the Luminary Cloud MCP server

  1. Check if your server is running:

    • Open your mcp.json file where servers are configured

    • Look for the "Running" status and click "Start" if needed

    • Alternatively, use Command Palette → MCP: List Servers

  2. Verify tools are active:

    • In Copilot, click the icon in the bottom right

    • You should see Luminary Cloud MCP tools in the dropdown list

    • Look for tools like get_simulation_setup_code, list_projects, sdk_instruction, etc.

"Request failed" error with too many tools

If you see an error about having more than 128 tools in your request:

  • Click the Copilot icon and deactivate some tools to get below the limit

VSCode won't show auth window on macOS

If VSCode can't display the authentication window and shows constant 401 errors:

  • Check System Preferences → Privacy & Security → Automation

  • Find "Visual Studio Code" in the list

  • Make sure "System Events" is toggled on

Cursor Issues

Red dot in the MCP server list

Don't worry—a red dot doesn't mean the MCP server is disconnected. Your tools will still work perfectly. You can either:

  • Leave it as is (everything will work fine)

  • Toggle the MCP server off and on again to get the green dot back

---

Having other issues? We're here to help—reach out to our support team with your specific error messages and we'll get you up and running quickly.

Did this answer your question?