mcpdata Request an invite

Connect your AI

One address, one key.

Add mcpdata to your assistant as a remote MCP server at https://[you].mcpdata.org/mcp with your API key. From Python, install the client and use the same key.

To try it first, use the public demo instead: address https://demo.mcpdata.org/mcp and this read-only key, public on purpose (it can only read the demo's synthetic sample data):

cdb_live_PkVoldus8qcE-hqffjajNZzQwiA7rTHMoR4RLSAxsCE

Claude Code

claude mcp add --transport http mcpdata https://[you].mcpdata.org/mcp --header "Authorization: Bearer $CDB_API_KEY"

Then /mcp in Claude Code shows it connected.

Cursor

In ~/.cursor/mcp.json (or .cursor/mcp.json in a project):

{
  "mcpServers": {
    "mcpdata": {
      "url": "https://[you].mcpdata.org/mcp",
      "headers": { "Authorization": "Bearer YOUR_KEY" }
    }
  }
}

VS Code

In .vscode/mcp.json; VS Code asks for the key once and keeps it:

{
  "inputs": [{ "type": "promptString", "id": "mcpdata-key", "description": "mcpdata API key", "password": true }],
  "servers": {
    "mcpdata": {
      "type": "http",
      "url": "https://[you].mcpdata.org/mcp",
      "headers": { "Authorization": "Bearer ${input:mcpdata-key}" }
    }
  }
}

Claude Desktop

Claude Desktop's config file starts local programs, so a remote server with a key goes through the mcp-remote bridge (it needs Node.js). In claude_desktop_config.json (Settings, Developer, Edit Config), then restart Claude Desktop:

{
  "mcpServers": {
    "mcpdata": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://[you].mcpdata.org/mcp", "--header", "Authorization:${AUTH}"],
      "env": { "AUTH": "Bearer YOUR_KEY" }
    }
  }
}

claude.ai and ChatGPT

Connecting these by signing in (OAuth) instead of a key is built and in testing with each app; this page will give the steps once they're verified.

Python and Jupyter

Install the client from your instance's help page, then use the same tables from notebooks and scripts:

pip install https://[you].mcpdata.org/help/download/cdb-0.3.4-py3-none-any.whl
import cdb
s = cdb.Session("https://[you].mcpdata.org")
print(s.execute_sql("SELECT * FROM SYSTEM.TABLES").to_dataframe())   # your tables, with their row counts

What your AI gets

Tools to explore (list_tables, describe_table), query (query), keep its own tables (create_table, append_rows, replace_rows), run and save views (list_views, query_view, create_view, drop_view) and build reports (create_report, run_report, schedule_report, list_reports, list_report_runs, delete_report), plus a schema catalogue and guides it reads before writing queries. The full list, generated from the server itself: MCP tools.

Your keys

Create a key per assistant in the accounts portal. Choose its role (Read, ReadWrite, Admin), when it expires, and which addresses may use it. Command-line tools can sign in through the browser instead of a pasted key (cdb-localdata login).

Answers about connecting