Inspect a local CSV from an MCP-compatible client

What the MCP server exposes

DataCello bundles an MCP stdio server with three tools: inspect_dataset, run_analysis, and show_analysis_result_in_app. This guide uses only inspect_dataset to return columns, inferred types, row count, and a small preview for a local CSV.

The current MCP server code does not include a DataCello Pro check, so this page is not labeled Pro. Any configuration or cost in the MCP client remains separate.

Prepare the connection

1. Install DataCello and make its bundled analytics engine available.
2. Add the DataCello MCP stdio command to the MCP client's configuration file.
3. Find the absolute path of the CSV.
4. List available tools and confirm inspect_dataset appears.

Configuration syntax differs by client. This is a local stdio process, not a DataCello HTTP server.

Inspect the CSV

Pass a non-empty CSV path to inspect_dataset. Use the returned schema and preview to identify possible targets, numeric features, and date fields before selecting an analysis.

tool: inspect_dataset
path: /Users/you/Documents/orders.csv

Use it safely

• Review the MCP client's data-handling policy because it mediates access to the local file.
• inspect_dataset reports structure; it does not repair the CSV.
• Client setup and permission screens are not part of the DataCello interface.
• Opening a structured result in the app uses the separate show_analysis_result_in_app tool.