How to Connect Claude to Live Data Sources in Under 2 Minutes

Claude can do more than answer questions from its training data. With MCP (Model Context Protocol) servers, you can give Claude access to live tools — including dataset search across 21+ platforms. Here's how to set it up.

Option 1: Claude.ai (hosted, no install)

This is the fastest path. No downloads, no configuration files.

1. Open Claude.ai and go to Settings 2. Navigate to the MCP or Integrations section 3. Add a new MCP connection with the URL: https://mcp.mobus.ai/mcp 4. Authenticate with your account 5. Start a new conversation and ask Claude to search for datasets

That's it. Claude now has access to tools like search_datasets, preview_dataset, check_license, and generate_citation.

Option 2: Claude Desktop (local config)

If you're using Claude Desktop, you'll add the MCP server to your configuration file.

1. Open your Claude Desktop config (usually claude_desktop_config.json) 2. Add the Mobus server to the mcpServers section 3. Restart Claude Desktop 4. The tools will appear in your conversation

Option 3: Self-host

For teams that want full control, you can run the Mobus MCP server locally:

1. Clone the repository from GitHub 2. Install dependencies 3. Configure your environment 4. Run the server and point Claude to your local endpoint

Self-hosting gives you control over caching, rate limits, and which platforms are enabled.

What you can do once connected

With the MCP server connected, Claude can execute 15 specialized tools:

Discovery — Search across Kaggle, Hugging Face, Zenodo, and 18 other platforms simultaneously. Filter by license, format, size, and recency.

Evaluation — Preview actual data rows, inspect schemas, check column types, and assess quality metrics without downloading anything.

Compliance — Verify licenses for commercial use, check terms of service, and flag restrictions.

Citation — Generate properly formatted BibTeX, APA, or Chicago citations directly from dataset metadata.

Monitoring — Set up alerts for new datasets matching your criteria on specific platforms.

Tips for effective use

Be specific in your prompts. Instead of "find climate data," try "find monthly temperature anomaly datasets from 2010-2024 with global coverage and a permissive license." The more context you provide, the more targeted the results.

You can also chain operations: "Search for EU air quality datasets, preview the top result, check its license, and generate a BibTeX citation" — all in a single conversation turn.