A fast, native Mac spreadsheet that automates with real Python — not VBA, not macros trapped in the cloud. Works fully offline. Your data never leaves your Mac. Buy once, yours forever.
FastSheet is a Mac app. Open fast-sheet.com on your Mac to download it.
Sales data with a live pivot table, and the Python macro editor beside the sheet.
Every spreadsheet has that one tedious job — reformat, recompute, re-flag, repeat. In FastSheet you write it once, in real Python, and run it over your cells. No VBA to relearn, no add-in to install, no cloud runtime.
The macro editor speaks the whole language — loops, functions, the standard library — over a simple Cell / Range object model. It runs instantly, on your machine, against the sheet in front of you.
Import math, datetime, re, json, the built-in numpy, pandas and scipy modules, or your own .py files — write the transform the way you'd write it anywhere else.
# Build the month-end summary and colour the misses red import datetime targets = {"EU": 50000, "US": 40000} today = datetime.date.today() for row in Range("A2:C50"): zone, sales = row["C"], row["B"] if sales < targets.get(zone, 0): row.fill("#ffd7d7") # missed target row["D"] = f"short by {targets[zone] - sales:,}" Cell("F1") = f"Closed {today:%b %d}"
Real Python over your cells — loops, functions, the whole standard library. The report you rebuild every week becomes one command.
400+ functions — statistics, matrices, distributions, pivots, charts. Opens and saves .xlsx with live formulas in both directions.
No account, no subscription, no cloud. It runs on your Mac and stays there — the only thing that ever leaves is a file you choose to share.
The parts of Excel people still open it for — in FastSheet, offline and scriptable in Python.
Like Excel's Visual Basic editor, beside the sheet: a project list, code tabs with colours and completion, output, F5 to run. Modules share their functions with every macro; your own .py files import too.
Draw a form — text boxes, lists, buttons — in the form designer, double-click a button for its code, and run it over the sheet. The forms are saved in the workbook.
Max, min or a target value by changing cells, with constraints, integer and binary cells. Simplex LP, GRG-style nonlinear, Nelder–Mead and evolutionary methods, and an answer report.
Bar, stacked and 100% columns, histogram, Pareto, box & whisker, waterfall, funnel, stock, treemap, bubble, 3D — with values under the cursor and zoom.
Click a value to filter a table or a PivotTable; ⌘-click for several. Pivots, sort and AutoFilter, conditional formats, data validation.
Rest the pointer on a marked cell: read, edit, reply or delete in a small card at its corner. Replies make a comment thread you can resolve.
Have Ollama on your Mac? FastSheet finds it by itself — no settings, and nothing leaves your Mac. With the free qwen2.5-coder:7b model it did 56 of 62 real sheet tasks right in our tests. Or use Claude with your own Anthropic key. No Ollama, no key: no AI, nothing sent.
Tools ▸ AI ▸ AI Assistant, beside the sheet: ask what is in the data, or ask for a change — "add a Revenue column", "sort by date", "copy the West rows to Sheet2". The AI's macro runs only when you click ▶ Run: FastSheet checks it first, sends any error back to the AI for a fix, and Undo takes a run back. No AI set up? Copy the macro guide into ChatGPT or Claude.
Classify, extract or summarise right in the grid. Answers are cached and saved with the workbook, so a recalculation — or opening the file again — asks nothing twice.
FastSheet=AI("Positive or negative?", B2)
fastsheet-mcp is an MCP server: Claude Desktop, Claude Code and other agents open, read and write workbooks, run macros and save .xlsx — with FastSheet's own engine.
claude mcp add fastsheet -- ~/bin/fastsheet-mcp
An incremental engine recomputes only what changed, so big sheets stay fluid — and when your data outgrows cells, FastSheet queries real database files in place.
A small native app — no Electron, no Java. Each sheet has a million rows; the grid draws only what you see and loads the rest as you scroll, and the incremental engine recomputes only the cells that depend on your edit — so a change lands in about five milliseconds. When everything must recompute, 100,000 formulas take about two seconds (about seven the first time, while each formula is translated).
Link a CSV or database file and query it live with DBSUM, DBLOOKUP, DBAVG, DBCOUNT — backed by a real SQLite engine, no import step. Edit the source, hit refresh, your sheet updates. A spreadsheet that talks to a database.
A big CSV opens as a Data Table: sort, filter and group 20 million rows on a 16 GB Mac. Excel stops at 1,048,576 rows, Numbers at 1,000,000.
| Same 1,000,000-row CSV | FastSheet | Numbers |
|---|---|---|
| Open the file | 0.3 s | 19 s |
| Sort by a column | 0.02 s | 15 s |
| Filter (region = North) | 0.02 s | 5 s, then ~1 min “Calculating…” |
| Memory in use | ~0.5 GB | ~2.7 GB |
| Most rows per table | 20 million+ | 1,000,000 |
Measured 17 Oct 2026 on one MacBook (Apple M5, 16 GB, macOS 26.6) with Numbers 14.5: a 65 MB CSV with 10 columns. FastSheet times are its Data Table engine through File ▸ Import CSV; Numbers was timed through its own window.
It won't replace Excel for everyone, and it doesn't try to. It's for people who want a fast native spreadsheet they can script — and own.
| FastSheet | Excel 365 | Numbers | Google Sheets | |
|---|---|---|---|---|
| Price | $19 once · first 100 | ~$100 / year | Free | Free |
| Scripting | Python | VBA | — | Apps Script (JS) |
| Works offline | Yes | Partial | Yes | No |
| Data stays local | Always | Cloud-linked | Yes | Cloud |
| Reads / writes .xlsx | Yes | Yes | Import/export | Import/export |
| Most rows | 1M per sheet · 20M+ in a Data Table | 1,048,576 | 1,000,000 | 10M cells |
Fourteen days free, fully unlocked, no card. Then one payment keeps it forever, with a year of updates — no renewals, no seats, no cloud login. The first 100 buyers pay $19 instead of $39; in return, you get my ear — email bugs and ideas straight to [email protected] and help shape v1.
Does it open my Excel files?
Yes — .xlsx in and out, with live formulas both directions. Edit a cell and dependent formulas recompute, in FastSheet or back in Excel.
Is my data uploaded anywhere?
Never by default. FastSheet works fully offline and your files stay on your Mac. It goes online only for the one-time licence unlock and, if you choose Claude, for AI. A local model (Ollama, found by itself when it runs on your Mac) keeps everything on your Mac; Claude sends what you ask to Anthropic, with your own key.
Is it a subscription?
No. Buy once. A year of updates is included, and the version you bought keeps working forever, offline.
What language are the macros?
Python — the real thing, with the standard library. Not VBA, not a cloud runtime. Runs instantly on your machine.
Which Macs does it run on?
Apple Silicon and Intel, as a native app. No Rosetta, no Java, no Electron bloat — one small download.
What if it's not for me?
The 14-day trial is fully unlocked — you'll know before you pay a cent. And there's a refund window on every purchase.
How much data can it hold?
Each sheet has 1,000,000 rows and 52 columns (A–AZ). A bigger CSV opens as a Data Table: 20 million rows sort, filter and group in well under a second on a 16 GB Mac, and formulas read it with DBSUM and friends.
How does it compare with Numbers?
Numbers is free and lovely for small tables. On the same 1,000,000-row CSV, FastSheet opened it in 0.3 s and sorted it in 0.02 s; Numbers took 19 s and 15 s. And FastSheet runs Python macros and keeps Excel files as Excel files.
macOS 12+ · Apple Silicon & Intel · ~25 MB · notarised by Apple