Excel MCP

Excel MCP lets Claude create, read, and manipulate Excel workbooks without requiring Excel to be installed. It's built on openpyxl, so you get proper .xlsx files with charts, pivot tables, formulae, and formatting. The server exposes tools for creating workbooks, adding sheets, writing data, inserting formulae, and building charts. I've used it to generate monthly financial reports from API data, and it handles the entire pipeline: Claude fetches the data, structures it into tables, adds SUM and AVERAGE formulae, then drops in a bar chart. The output is a proper Excel file, not a CSV pretending to be one. The real win here is batch generation. If you're producing dozens of similar reports with different data sets, Claude can loop through them without you touching Excel. It's also useful for data wrangling tasks where you'd normally open a spreadsheet, apply a formula column, then export. Now Claude does it in one pass. Chart support is solid for common types (bar, line, pie, scatter), and pivot tables work as long as your data is already clean and structured. Quirks: it's Python-based, so you need uvx or a Python environment. The server doesn't validate formulae before writing them, so if you ask Claude to insert a broken formula, you'll only find out when you open the file. Pivot table creation assumes you know the field names and aggregation types upfront. It won't guess. Also, this is file-based: it writes to disk, so you'll need to specify output paths. No in-memory streaming. Skip this if you're just exporting CSVs or doing simple tabular work. A basic file-write MCP or even stdout redirection is faster. But if you're generating actual Excel artefacts with charts and formulae, or you're automating report templates, this is the right tool. It's not trying to be a full Excel replacement, it's trying to let Claude script the boring parts of spreadsheet work, and it does that well.