ai.baselight/baselight

Official9 tools

Query Baselight's public catalog of 70,000+ datasets — finance, demographics, sports, and more.

Query public catalog of 70,000+ datasets covering finance, demographics, sports, and more.

Captured live from the server via tools/list.

baselight_ping

Simple ping test to verify MCP server is responding

No parameters.

baselight_search_catalog

Search the catalog for datasets using a text query and filters. Datasets in Baselight have the following format: @username.dataset. Datasets can be public or private — you can search and use all public datasets as well as the user's private datasets. This is typically the first step in the discovery workflow.

Parameters (3)

querystringrequired
The search query string, used for semantic search
categoryarray
The dataset category to filter by
limitnumber
Maximum number of datasets to return (1-20, default 10)

baselight_search_tables

Search for tables using a text query and filters. Tables in Baselight have the following format: @username.dataset.table. Tables are grouped into datasets which can be public or private — you can search and use all public datasets as well as the user's private datasets. Search for tables directly when you are unable to find relevant datasets.

Parameters (3)

querystringrequired
The search query string, used for semantic search
categoryarray
The dataset category to filter by
limitnumber
Maximum number of tables to return (1-20, default 10)

baselight_get_dataset_metadata

Retrieve detailed schema and metadata for a specific dataset using Baselight format @username.dataset. Use this after discovering datasets to understand their structure before querying. Tables within datasets follow the format @username.dataset.table (always double-quoted identifiers in SQL).

Parameters (1)

idstringrequired
The identifier of the dataset to inspect, should be something like @user.dataset

baselight_get_dataset_tables

Retrieve a paginated list of tables from a specific dataset using Baselight format @username.dataset. Use this tool to browse datasets with many tables or to search for specific tables within a dataset. Each page returns up to 100 tables with metadata.

Parameters (3)

idstringrequired
The identifier of the dataset to browse, should be something like @user.dataset
querystring
Optional text search query to filter tables by name or title. If omitted, returns all tables.
pagenumber
Page number to retrieve (starts at 1). Each page contains up to 100 tables.

baselight_get_table_metadata

Retrieve detailed schema and metadata for a specific table using Baselight format @username.dataset.table. Use this to understand table structure, column types, and constraints before writing SQL queries. Tables must be referenced in SQL with double quotes.

Parameters (1)

idstringrequired
The full identifier of the table to inspect, should be something like @user.dataset.table

baselight_get_user

Retrieve detailed information about a Baselight user including their profile bio, website, and public datasets. Use this to understand who owns datasets you're interested in.

Parameters (1)

usernamestringrequired
The username of the user to inspect

baselight_sdk_query_execute

Execute a SQL query on Baselight and wait for results (up to 1 minute). The query executes and returns the first 100 rows upon completion, or info about a pending query that needs more time. Use DuckDB syntax only, table format "@username.dataset.table" (double-quoted), SELECT queries only (no DDL/DML), no semicolon terminators, use LIMIT not TOP. If query is still PENDING, use `sdk-get-results` to continue polling. If totalResults > returned rows, use `sdk-get-results` with offset to paginate.

Parameters (1)

sqlstringrequired
The SQL query to execute. table identifiers should be wrapped in double quotes, like "@user.dataset.table". Only SELECT queries are allowed.

baselight_sdk_get_results

Retrieve results from a previously executed SDK job using the resultId from `sdk-query-execute`. If the query is complete, returns results immediately. If still pending, polls for up to 1 more minute. Use this after `sdk-query-execute` returns PENDING status.

Parameters (3)

jobIdstringrequired
The result ID of the executed query (from `sdk-query-execute`).
limitnumber
Number of rows to return per page (max 100). Default is 100.
offsetnumber
Row offset for pagination. Use to fetch subsequent pages (e.g., offset=100 for page 2).

Install

claude_desktop_config.json

{
  "mcpServers": {
    "baselight": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://api.baselight.app/mcp"
      ]
    }
  }
}

Desktop config is stdio-only; this bridges via mcp-remote. Native remote: Settings > Connectors.