Summary
The local fallback is the part worth noting: most model-backed tools simply fail when the upstream does, and this one degrades instead.
A generator that depends entirely on a hosted model has a single point of failure that shows up as an error at the worst moment. Having a local parse path means a simpler request still produces a file when the API call does not land. Beyond that this is deliberately one tool doing one thing — describe, receive a CSV — and the value is in skipping the step where you specify columns you have not decided on yet. Documentation and tool text are in Chinese.
What it is
A single-tool CSV generator: it takes a natural-language description, derives the column structure and the rows, and writes a file with a descriptive name.
What you get
- generate-csv produces the file from a description, extracting both the column structure and the data content rather than requiring a schema.
- The output filename is generated to describe the contents instead of being left to the caller.
- The delimiter is configurable, so the output can be something other than comma-separated.
- There is a local parsing fallback for when the Gemini call fails, so a service outage degrades rather than blocks.
Requirements
A Google Gemini API key, passed as a command-line argument at launch.
Setup effort
One command — npx -y structurize-mcp