Monthly Report (Apple Numbers)

Build a readable monthly summary on top of a live Numbers spreadsheet using the Digits MCP tools (numbers_*). The source data is whatever rows the user already has (transactions, sales, hours, etc.); your job is to aggregate it faithfully — never invent numbers, always read before writing.

Workflow

1. Locate the data. Call numbers_list_documents. If empty, ask for the file path and numbers_open_document. Use numbers_list_sheets and numbers_list_tables to find the source table and confirm its dimensions.
2. Read the source. Call numbers_read_table with include_formulas: true so you don't mistake a formula cell for a typed constant. Identify the key columns: a date/period column, one or more category/label columns, and the numeric column(s) to aggregate.
3. Aggregate in your head, not in guesses. Compute the rollups you intend to write (per-category totals, grand total, count, average, and — if a prior period is present — the month-over-month delta and % change). Do the arithmetic from the values you actually read.
4. Decide where the summary lives. Prefer a dedicated summary sheet so you never clobber source rows. If one doesn't exist, offer to create it with numbers_add_sheet (new sheets start blank — see Limitations). Lay out a small table: a header row, one row per category, then a Total row.
5. Write the summary in one batch. Use a single numbers_set_cells call for the whole summary. For the Total row and any derived figures, write formulas (=SUM(B2:B9)), not pre-computed constants, so the report stays live if the user edits a number later. (Digits has no transaction API — one batched call is the closest thing to atomic, so build the whole block at once rather than dribbling cells in.)
6. Verify. Re-read the summary table and confirm the formula results match the totals you computed in step 3. If they disagree, the cell references are wrong — fix them before reporting success.
7. Export only with permission. If the user wants a shareable artifact, confirm the path and format, then numbers_export (pdf, excel, or csv). Saving the .numbers file itself is a separate, explicit step (numbers_save_document / numbers_save_as) — ask first.

Rules

• Read before you write; the source rows may have changed since last time.
• Aggregate only the values you read — no estimates, no "roughly."
• Put the summary on its own sheet/table; never overwrite source data.
• Derived cells are formulas, not constants, so the report recomputes.
• One numbers_set_cells batch per summary block, not many small writes.
• Numbers are plain (1234.5), never formatted strings ("$1,234.50").
• State your assumptions about which column is the period and which is the amount, and let the user correct you before you write.