Search and preview a dataset through a public MCP endpoint
At a glance
A short Agent Framework script searches a public dataset catalog and previews sample rows before a developer decides whether to download. This workflow is listed under MCP, dataset discovery, Python, agent tools.
- Agents / tools
- Microsoft Agent Framework, MCPStreamableHTTPTool, DataBazaar public MCP, Python, uv
SourceBuilding a tool for these workflows? See weekly rates.
What they did
1) Set up Python 3.12 and save the example as discover.py; run it with uv run --python 3.12 --with agent-framework-core==1.21.0 --with mcp==1.30.0 python discover.py. The script connects MCPStreamableHTTPTool to https://api.databazaar.io/mcp/public, disables prompt loading, and allows only search_datasets, get_dataset, and preview_sample. 2) Call search_datasets with the query wireless earbuds and limit 1, then pass the returned dataset ID to preview_sample. 3) Print the listing and sample, inspect the fields and rows, and decide whether the dataset fits before requesting a full download.
How it went
The post says the example was checked against the live endpoint on October 10: it returned a wireless-earbuds dataset and a CSV sample with 10 rows, including product_name, price, and retailer fields.
Worth knowing
Searching and previewing the public catalog require no API key, but full downloads require a DataBazaar account. A sample preview shows format and example rows, not the accuracy or suitability of the full dataset. The author notes that the post used AI assistance and that the direct tool calls were checked live.
