Excel Parser

Tool: parse_xlsx

Downloads an Excel file from a URL and returns its sheets as Markdown tables.

Arguments

Every sheet is parsed, whole; there is no sheet or row argument.

Output

For each sheet, in order: a --- Sheet: <name> --- marker, then the sheet as a Markdown table with its column headers in the first row — or [Empty Sheet].

If the download or parsing fails, the call fails — Error parsing XLSX: … with the reason — and a Tool step that made it fails with that reason; it is never the tool's output.

Only URLs, and which ones

The file is fetched over HTTP from the SynapseOrch server, so it cannot be a local path or a Vault file. A host on your own network — loopback, private or link-local addresses, including where a redirect leads — is refused unless it is listed in EGRESS_ALLOW_HOSTS. A download follows at most 5 redirects, must finish within 60 seconds, and may be at most 25 MB.

Use cases

Parse the sales data at https://example.com/exports/q3-sales.xlsx and identify the top 10 accounts by revenue.

Read the employee list at [URL] and find all employees who joined after January 2024.

Analyse the inventory spreadsheet and flag any items with stock below 10.

Working with large spreadsheets

For spreadsheets with many rows, combine the Excel parser with the sandbox tool:

1. Agent parses the Excel file to understand the structure
2. Agent writes a pandas script to process the full dataset
3. Sandbox executes the script and returns the results

Look at the structure of the spreadsheet at [URL], then write a Python script to compute the full analysis and run it in the sandbox.

Supported formats

• .xlsx (Excel 2007+) — full support
• .xls (Excel 97-2003) — not supported (convert to .xlsx first)
• .csv — not supported (agents can read CSVs directly via vault or sandbox)