Baselight MCP Server for Claude

Overview

Baselight provides a hosted Model Context Protocol server that exposes a catalog of more than 70,000 public datasets. After OAuth or x-api-key authentication, Claude and other MCP clients can search the catalog, inspect dataset metadata, and run queries without switching tools.

The integration is documented in the BaselightDB/mcp repository and hosted at https://api.baselight.app/mcp.

Features

• Search a catalog of 70,000+ public datasets by name, topic, and tags.
• Inspect dataset schemas, descriptions, and freshness metadata.
• Run queries against accessible public datasets through MCP tools.
• OAuth or x-api-key header authentication.
• Hosted remote HTTP endpoint at api.baselight.app.

Use Cases

• Discover open datasets for a machine learning benchmark.
• Compare schema and row counts before importing a public table.
• Answer ad hoc analytical questions against curated public data.
• Find demographic or economic datasets for a research brief.
• Prototype dashboards by querying sample rows in chat.

Installation

Claude (Connectors)

1. Add a custom connector under Settings → Connectors.
2. Enter the MCP URL: https://api.baselight.app/mcp.
3. Complete OAuth sign-in or configure the x-api-key header.
4. Enable the connector in your next conversation.

Claude Code

claude mcp add --transport http baselight https://api.baselight.app/mcp
claude mcp list

Add x-api-key in your MCP config when not using OAuth.

Other MCP clients

Point a remote HTTP connector at https://api.baselight.app/mcp and complete OAuth or API key setup.

Configuration

{
  "mcpServers": {
    "baselight": {
      "url": "https://api.baselight.app/mcp",
      "type": "http",
      "headers": {
        "x-api-key": "YOUR_BASELIGHT_API_KEY"
      }
    }
  }
}

Omit headers when authenticating through OAuth in a connector UI.

Examples

Search the catalog

Search Baselight for public datasets about global temperature anomalies and list the top five matches.

Inspect schema

Show the schema and row count for the most recent COVID cases dataset in Baselight.

Sample query

Query the first 10 rows of a US census population dataset and summarise the columns.

Security

• Review dataset licenses before redistributing or commercial use.
• Limit query scope to avoid downloading large sensitive extracts unintentionally.
• Revoke API keys and OAuth tokens when offboarding users.
• Treat returned rows as potentially sensitive even when labeled public.

Troubleshooting

OAuth or 401 errors

Re-authorise the connector or verify the x-api-key value in Baselight account settings.

Empty search results

Broaden search terms or remove restrictive tag filters; confirm the dataset is in the public catalog.

Query timeouts

Add filters or reduce row limits; very wide tables may exceed default timeouts.

Schema mismatch

Community datasets may change schema; refresh metadata before reusing saved queries.