Summary
Spreadsheet work answered in the conversation, charts included.
The usual load, sample, filter and describe loop is here, and so is the part most data servers stop short of: it renders the heatmap or the distribution plot rather than describing what you would see.
What it is
An MCP server built on VisiData that treats a data file as something you query in conversation — load it, sample it, filter, sort, convert between formats, and render correlation heatmaps and distribution plots.
What you get
- A data file loaded and inspected, in the formats VisiData reads
- A preview of the rows, with the row count you ask for
- Column types and statistics across the file, or detail for one column
- Rows filtered by equals, contains, greater than or less than, and sorted by any column
- Conversion between CSV, JSON, Excel and the other supported formats
- Correlation heatmaps, distribution plots (histogram, box, violin, kde) and scatter, line or bar graphs
- A comma-separated skills column split into individual skills with one-hot encoding, and salary and frequency broken down by location
Requirements
A recent Python. Install with npm install -g @moeloubani/visidata-mcp@beta or pip install visidata-mcp, then point the client at the visidata-mcp command. No account, no key — it reads files you already have.
Setup effort
One command — npm install -g @moeloubani/visidata-mcp@beta