Stars

MCP Server

PortalJS Cloud exposes your portal's data through a Model Context Protocol (MCP) server. Connect it to any MCP-compatible AI chatbot (Claude, Cursor, etc.) and explore your open data with natural language.

Find your MCP URL

  1. In the sidebar, click MCP server.

    placeholder: MCP server page

  2. Your MCP endpoint is shown in the input box. It looks like:

    https://mcp.portaljs.com/@<your-org-name>/sse

  3. Click Copy to put it on your clipboard.

    placeholder: copy MCP URL

Connect to a chatbot

The exact steps depend on the chatbot. In general:

  1. Open your chatbot's MCP / integrations / connectors settings.
  2. Add a new MCP server, pasting the URL you copied.
  3. Save and (re)start the chat session.
  4. Ask the chatbot questions about your data — for example, "What datasets do you know about? Show me the top 5 rows of the budget dataset."

For a walkthrough, see the demo video embedded on the MCP server page.

placeholder: MCP demo video

Example: connect with Claude (web)

  1. Copy your MCP URL from the MCP server page (see Find your MCP URL).

  2. In Claude on the web, open Customize → Connectors.

  3. Click Add custom connector.

    placeholder: Claude add custom connector

  4. Give it a name (for example, My Data Portal) and paste your MCP URL into the Remote MCP server URL field, then confirm.

    placeholder: Claude connector URL field placeholder: Claude connector URL field

  5. Start a new chat. Open the tools/attachments menu and make sure your connector is enabled.

  6. Ask a question about your data, for example: "What datasets are in my portal? Show me the first 5 rows of the dataset." Claude calls the MCP server and answers using your portal's data.

    placeholder: Claude querying portal data

What the chatbot can do

Through the MCP server, the chatbot can:

  • List datasets and resources in your portal.
  • Read resource contents (CSV, JSON, etc.).
  • Answer questions and produce summaries grounded in your data.

Private datasets are not exposed via the MCP server.

Troubleshooting

  • Chatbot can't connect — confirm you pasted the full URL including the /sse suffix.
  • Chatbot sees no datasets — make sure at least one dataset is marked Public. See Datasets → Visibility.