Skip to main content

​
The Veydra Intelligence Layer

Connecting Veydra to your AI takes less than two minutes. Once connected, your AI assistant can run system dynamics simulations, analyze causal loops, and provide decision support directly in your chat window.
You’ll need a Veydra API key for most integrations. Get yours at veydra.io/settings/api-keys.

​
Claude Desktop

Claude is currently the most robust host for MCP. It uses a configuration file to manage connections.
1

Open Settings

Click your profile name in the bottom-left corner of Claude Desktop and select Settings.
2

Developer Tab

Navigate to the Developer tab in the sidebar.
3

Edit Config

Click the Edit Config button. This will open your claude_desktop_config.json file.
4

Add Veydra

Paste the following into the mcpServers object:
"veydra": {
  "url": "https://api.veydra.io/mcp",
  "headers": { "Authorization": "Bearer YOUR_API_KEY" }
}
5

Restart

Fully quit and restart Claude. You will see a 🧩 icon in the chat bar indicating Veydra is active.

​
Cursor & VS Code

Cursor provides a native GUI for adding MCP servers without touching JSON files.
1

Open Settings

Press Cmd + Shift + J (Mac) or Ctrl + Shift + J (Windows) to open Cursor Settings.
2

Navigate to MCP

Scroll down to the MCP section under General.
3

Add New Server

Click the + Add New MCP Server button.
4

Fill the Form

Configure the server with these values:
FieldValue
NameVeydra
Typesse (or HTTP)
URLhttps://api.veydra.io/mcp
5

Save

Click Save. Cursor will immediately “handshake” with Veydra and list available tools.

​
ChatGPT (Enterprise & Plus)

OpenAI supports MCP through “Connectors” (Developer Mode required).
1

Enable Developer Mode

Go to Settings → Connectors → Advanced and toggle Developer Mode to ON.
2

Create Connector

Go to Settings → Connectors and click the Create button.
3

Configure URL

Fill in the connector details:
FieldValue
NameVeydra Simulation
Connector URLhttps://api.veydra.io/mcp
4

Authentication

Set authentication to Bearer Token and paste your Veydra API Key.
5

Verify

ChatGPT will ping the endpoint and show you a list of tools like list_models, run_simulation, or get_parameters.

​
Grok (xAI)

Grok 3+ supports remote MCP tools via their “Plugin” architecture.
1

Plugin Settings

In the Grok interface, click the Plugins icon (usually a ⚡ or 🔌 icon).
2

Custom Tool

Select Add Custom Remote Tool.
3

Manifest URL

Enter https://api.veydra.io/.well-known/mcp.json. Grok will automatically parse your server capabilities.
4

Authorize

Confirm the permissions Grok requests to access the Veydra simulation engine.

​
Gemini CLI

Google’s Gemini CLI supports MCP servers natively. 
gemini mcp add veydra -- npx @veydra-io/mcp-server
Or configure with a URL: 
{
  "mcpServers": {
    "veydra": {
      "url": "https://api.veydra.io/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

​
Local-First Setup (For Researchers)

If you’re a data scientist working with sensitive local data, we recommend the stdio version. This ensures your data never leaves your infrastructure.
The local version requires Node.js 18+ installed on your system.

​
Installation

npm install -g @veydra-io/mcp-server

​
Usage with Claude Desktop

Add to your claude_desktop_config.json: 
"veydra": {
  "command": "npx",
  "args": ["@veydra-io/mcp-server"],
  "env": {
    "VEYDRA_API_KEY": "your-api-key"
  }
}

​
Environment Variables

VariableDescriptionRequired
VEYDRA_API_KEYYour Veydra API keyYes
VEYDRA_API_URLCustom API URL (default: https://api.veydra.io)No

​
Available Tools

Once connected, your AI assistant has access to these tools:
ToolDescription
list_modelsList available simulation models
get_modelGet model details and structure
get_parametersGet adjustable model parameters
run_simulationExecute simulation with parameters
ask_modelChat with AI about a model

​
Troubleshooting

  1. Verify your API key is valid at veydra.io/settings/api-keys
  2. Ensure the URL is exactly https://api.veydra.io/mcp
  3. Check that your firewall allows outbound HTTPS connections
  1. Restart your AI client completely (not just refresh)
  2. Check that the MCP server is enabled in your client settings
  3. Verify authentication is configured correctly
  1. Ensure your API key starts with vk_
  2. Check that the Authorization header format is Bearer YOUR_API_KEY
  3. Generate a new API key if the current one may be compromised

​
Need Help?

API Reference

Explore the full API documentation

Support

Contact our support team