1. Show
Provide the original structured CSV or XLSX table and the finished result.
DETERMINISTIC · LOCAL · REVIEWABLE
Show the change.
Reuse the
pipeline.
Turn supported Before-and-After spreadsheet transformations into verified, reusable local pipelines. No LLM-generated transformation logic. No silent guesses.
Local-first · Exact example reproduction · Reviewable Python and SQLite SQL · Benchmarked up to 3M rows
Weekly orders
verified against example
- 01Keep paid ordersstatus equals “paid”
- 02Calculate net totalunit price × quantity
- 03Group by departmentsum net total
- 04Sort the reporthighest total first
Same input contract. Deterministic output. Reusable locally.
From example to
repeatable work.
FIXIT searches only within its declared transformation grammar. A candidate is accepted only when it reproduces the supplied result exactly. If several candidates fit, you choose the intended interpretation.
2. Confirm
Review the inferred steps. When an example is ambiguous, FIXIT asks instead of silently guessing.
3. Reuse
Validate on a representative new file, save the tool, and apply the same program again.
One core. Two
ways to use it.
The desktop product serves recurring spreadsheet work. The same deterministic contract is available to engineering teams through the Engine v2 SDK and generated code.
DESKTOP
Professionals and teams
Turn recurring spreadsheet preparation into local tools without writing code.
- Before-and-After teaching workflow
- Local files and saved tools
- CSV and XLSX input and output
- Explicit questions when intent is ambiguous
- Windows and Linux desktop builds available; macOS packaging planned
ENGINE V2
Enterprise and infrastructure
Embed inspectable transformation programs into controlled data workflows. Generated programs remain independent of the teaching interface.
- Versioned, typed transformation IR
- Bounded compositional planning
- Safe two-table inner and left joins
- Portable, human-readable Python/Pandas and schema-bound SQLite SQL
- Behavioral-equivalence checks within the supported contract
- Stable local SDK/API and deterministic fingerprints
Desktop downloads.
FIXIT runs locally on your computer. Windows and Linux builds are available now; macOS packaging is planned next.
WindowsAvailable
FIXIT Data Engine for Windows
Install the desktop app and run spreadsheet transformations locally on your PC.
Download for WindowsWindows 10/11 · 64-bit
macOSComing soon
FIXIT Data Engine for macOS
Native packaging is planned after the Windows release is finalized.
LinuxAvailable
FIXIT Data Engine for Linux
Download the x86_64 AppImage and run FIXIT locally on your Linux desktop.
Download for LinuxLinux x86_64 · AppImage
After downloading
Open a terminal in the folder containing the file and run:
chmod +x FIXIT-Data-Engine-x86_64.AppImage
./FIXIT-Data-Engine-x86_64.AppImage
A clear contract,
not an open-
ended promise.
FIXIT is designed to be predictable. It accepts transformations inside its current scope and refuses examples it cannot reproduce reliably.
Supported today
- Structured CSV and XLSX tables
- Cleaning and fixed-value missing-data filling
- Single-condition filtering
- Addition, subtraction, multiplication and division
- Rounding, rename, select and reorder
- Stable sorting and duplicate removal
- Grouping with sum, mean, min, max and count
- Tested one-table compositions
- Safe two-table inner and left joins in Engine v2
Bounded by design
FIXIT is a deterministic engine for declared structured transformations—not a general Excel automation system or a general-purpose AI agent.
- Every accepted program must stay inside the supported transformation grammar
- Unsupported and ambiguous cases are surfaced explicitly rather than guessed
- Excel formulas, macros, pivots, arbitrary chains and unrestricted multi-table workflows remain outside the current contract
What “verified” means: exact reproduction proves that a candidate explains the supplied example. It does not prove the user’s unstated intent for every future dataset. Always validate a learned pipeline on representative new data before production use.
3M
rows in the largest current benchmark scenario—not a universal performance guarantee.
Measured against independently
produced expected results.
Current benchmarks cover cleaning, filling, filtering, arithmetic, rounding, deduplication, stable sorting, grouping, large lookup joins, and equivalent execution through the native engine, generated Python and generated SQLite SQL. Actual speed and memory use depend on hardware, data shape and selected operations.
Plain-language
policies.
These summaries describe the private-preview product. Final commercial terms may be updated before general availability.
Privacy
FIXIT Data Engine Desktop processes spreadsheet files locally. Files and saved transformation tools are not intentionally uploaded by the desktop application. Purchase and billing information is processed separately by the checkout provider. This website’s hosting provider may process ordinary technical request data required to serve the page.
Product terms
Private-preview terms apply during controlled evaluation. Final commercial terms may change before general availability.
Refunds
Refund terms will be published with public checkout before general availability.
Built locally.
Available by preview.
FIXIT Data Engine is currently being prepared for commercial release. Windows and Linux desktop builds and the enterprise Engine/SDK are available for controlled evaluation; macOS packaging will follow.