PARQUET MCP

Query Parquet data from your MCP client.

Import an Apache Parquet file through Data Connector, then give a compatible MCP client read-only access to the resulting managed table.

1. Import your Parquet file

Sign in at dataconnector.ai, choose a workspace, and upload the file. Data Connector sends the upload through its API, validates the file, reads supported types from its embedded schema, and creates a queryable table.

2. Add Data Connector to your client

Use the hosted remote MCP endpoint:

https://dataconnector.ai/api/mcp

Complete OAuth and authorize the workspace containing your imported table.

3. Explore the data

Try: List my tables, describe the imported Parquet table, then summarize its date range and row count. Data Connector exposes bounded tools for discovery, schema inspection, and read-only SQL.

Managed after import

Your MCP client does not receive the original upload URL or object-store location. It queries the managed table through Data Connector's workspace permissions and API.

Frequently asked questions

Can an MCP client query a Parquet file?

Yes. Import the Parquet file through Data Connector, then authorize its workspace from any compatible remote MCP client.

Does Data Connector preserve Parquet column types?

Data Connector reads supported Arrow column types from the Parquet schema when it creates the managed table.

Does my MCP client receive the Parquet file URL?

No. It queries the managed table through Data Connector and does not receive the original upload location.

Import Parquet Set up MCP