Publishing MCP Servers

LangDB already offers a set of managed MCP servers you can use out-of-the-box, but you can also publish your own custom MCP server by linking a GitHub repo or configuring it manually. Once deployed, your custom MCP server can also be added to a Virtual MCP Server in LangDB, allowing you to group multiple tools together under one interface.

Github Auto Setup

Paste your GitHub repo URL, and LangDB will attempt to generate the MCP server configuration automatically.

1. Open the "Publish MCP Server" dialog
2. Enter the GitHub Repository URL
3. LangDB inspects the repo (e.g., package.json) and pre-fills:
   • command
   • args
   • env structure

You can edit or override these values before publishing.

Manual Configuration

If you prefer full control, enable Manual configuration and provide the required fields:

General Info

Field	Required	Description
Name	✅	Unique identifier for your MCP server
Logo URL	✅	PNG/SVG URL (128×128 preferred)
Description	⬜	What your MCP server does
Visibility	✅	Public or Private
Categories	✅	Select or define categories

Configuration JSON

{
  "command": "npx",
  "args": [
    "-y",
    "your-package-name"
  ],
  "env": {
    "API_KEY": "{{api_key}}"
  }
}

info

You can use either npx or uvx as the command depending on your MCP setup.

• Wrap variable names using {{...}}
• These placeholders must match environment variables defined below

Environment Variables

Define each required variable:

Name	Type	Description	Required
api_key	String	API key passed to your MCP tool	false

Deploying Your MCP Server

1. Make sure all required fields are completed, including configuration and environment variables.
2. Resolve any validation warnings highlighted in the interface.
3. Click Publish to deploy.

Once published, your MCP server will be accessible via LangDB—privately within your project or publicly if selected—and ready to power tool calls in any compatible LLM setup.

You can also add this newly deployed MCP server into a Virtual MCP Server in LangDB to combine it with other tools and provide a unified interface to your LLMs.

---

Configuration Notes

• You can insert variables in both args and env blocks using {{...}}
• Automatic GitHub mode can be used as a starting point and then switched to manual for advanced edits