Using the docs with AI assistants

How to connect this documentation to Claude, ChatGPT, Cursor or VS Code over MCP, so the assistant answers from the guide.

Connect this documentation to your AI assistant — Claude, ChatGPT, Copilot, Cursor or another. Then, when you ask about the portal, the assistant looks the answer up in this guide and links to the page.

MCP server addressStreamable HTTP

https://guides.dataportal.gov.lt/api/mcp

No sign-in. The server is public; no keys needed.
Read-only. The assistant searches and reads, and changes nothing.
Always current. The same content you see on the site.

What is MCP?

MCP (Model Context Protocol) is an open standard that AI assistants use to connect to outside sources and tools. Anthropic introduced it in 2024, and most major AI tools now support it. The name is worth unpacking:

Model

The language model — the assistant's “brain”. It knows a lot, but only what it learned up to a certain date.

Context

The information the model sees while answering. The better the context, the better the answer.

Protocol

Shared rules for how the model talks to an outside source: what to ask and what answer to expect.

MCP is often compared to a USB-C port: one plug fits many devices. Likewise, this documentation is connected once, and any assistant that supports MCP can use it.

Why does it matter? Without a source to check, an assistant answers from memory. The portal may have changed since the model was trained, and it may simply invent a button label or an API parameter. With MCP, the assistant checks this guide first and backs its answer with a link you can open yourself:

You ask
“How do I download only the filtered rows?”
The assistant searches
It searches this documentation over MCP.
It reads the page
It gets the current text of the guide.
It answers with a link
The answer is based on the guide — you can check it.

How to connect

On claude.ai or in the Claude app:

Open Settings → Connectors and click Add custom connector.

Enter a name (for example, LDP documentation) and the server address https://guides.dataportal.gov.lt/api/mcp. Leave the authentication fields empty.

Click Add. In a chat, turn the connector on from the + button or the tools menu.

Whether you can add custom connectors depends on your Claude plan and your organisation's settings.

In a terminal:

claude mcp add --transport http ldp-docs https://guides.dataportal.gov.lt/api/mcp

Check it with claude mcp list, or with /mcp inside a session.

Open Settings → Apps & Connectors → Advanced settings and turn on Developer mode.

Back in Apps & Connectors, click Create, enter a name and the server address, and choose No authentication.

In a chat, pick the connector from the + button.

Developer mode is not available on every ChatGPT plan.

In .cursor/mcp.json (for one project) or ~/.cursor/mcp.json (for all projects):

{
  "mcpServers": {
    "ldp-docs": {
      "url": "https://guides.dataportal.gov.lt/api/mcp"
    }
  }
}

Then check under Cursor Settings → MCP that the server is enabled.

With GitHub Copilot — in .vscode/mcp.json:

{
  "servers": {
    "ldp-docs": {
      "type": "http",
      "url": "https://guides.dataportal.gov.lt/api/mcp"
    }
  }
}

Or run MCP: Add Server from the command palette and enter the address. The tools are available in Copilot Chat's agent mode.

Any application that supports MCP over Streamable HTTP works — give it the server address.

If an application only supports local (stdio) servers, use the mcp-remote bridge (requires Node.js):

{
  "mcpServers": {
    "ldp-docs": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://guides.dataportal.gov.lt/api/mcp"]
    }
  }
}

What to ask

Once connected, just ask — the assistant decides when to look things up in the documentation:

How do I download only the filtered rows as XLSX from dataportal.gov.lt?
What does a dash mean in the data table?
How do I find the next releases in the release calendar and add a date to my own calendar?
Write Python that fetches every row of a table through the API, not just the first page.

Languages

The documentation search covers both the Lithuanian and the English text of every page and answers in the language you ask in — even when your question uses a term from the other language, such as a button label from the Lithuanian portal. The portal's official statistics are published in both Lithuanian and English, so you can search for and download the data in whichever language suits you.

The assistant's tools

The server gives the assistant three tools. They cover both the user guide and the Data API documentation, in both languages.

search_docs

Searches the documentation — the same search as on the site, over both languages; returns pages in the language of the question.

get_page

Reads a whole page — the text, tables and the descriptions of screenshots.

list_pages

Lists every page, in the same order as the menu.

Privacy

The server needs no sign-in and sets no cookies. It only reads the public documentation.

What reaches the server is only what the assistant puts into a tool call — the search words or the page address — not your conversation.
What is kept: when a search or page lookup finds nothing, its words and language are logged, with nothing about who asked, so that missing topics can be added to the documentation.
As with any page on this site, the servers keep standard technical access logs.

Your conversation itself is handled by your AI assistant's provider, under its terms. More in the portal's privacy policy.

Without MCP

If your tool does not support MCP, you can still hand it the documentation:

One page

The Copy Markdown button at the top of the page. Or add .mdx to a page's address.

Table of contents

llms.txt — every page with a short description and a link.

All in one file

llms-full.txt — the whole documentation in English. Also in Lithuanian.

Tips & tricks

Want to try the server yourself? Run npx @modelcontextprotocol/inspector, choose Streamable HTTP, enter the server address and call the tools directly.
The assistant doesn't use it? Ask explicitly: “search the dataportal documentation”.