v2 · desktop builds on GitHub Releases · not code-signed yet

Do the cleanup once. Re-run it forever.

Unified Data Studio is a desktop app for people who live in other people's spreadsheets. Chain formula steps into a workflow, watch every row change as you build it, then run the same workflow over millions of rows. Your files never leave your machine.

Download for your machine See it run

customers_q3 · 3 files joined 8 of 47,318 rows

Workflow

A real pipeline. Click any step to jump to it.

Capabilities

Seven things it does, in the order you'd do them.

Files on disk, warehouses when you need them — nothing here needs a second tool or a person who knows Python.

01 — BRING IT IN

Several files become one dataset

Load CSV, Excel and JSON side by side and work across them as a single table. Shared headers line up automatically; pick which sheets and columns to pull in rather than loading a workbook you only need three columns from.

·Selective loading — choose sheets and columns before anything is read

·No row ceiling — files are processed in full, not truncated to a preview

·GB-scale files are the design target, not the edge case

CSVcustomers_q3.csv47,318 rows

XLSXregions.xlsx2 sheets

JSONsignup_events.json1.4 GB

merge on shared keys

One working dataset

customer_idfirstlastemail signupspendregion_coderegionsource_file

02 — CONNECT IT

Warehouses without pasting secrets into every workbook

Org connectors put ClickHouse and Databricks behind the same step model as a file import. An Admin registers the credential once; editors and viewers pick the connector by name and run — they never see the password or token.

·ClickHouse · Databricks first — more warehouses on the roadmap

·Secrets stay sealed — members get a handle, not the credential

·Same workflow chain — warehouse rows join file steps like any other input

Org Admin

Registers once

Credential sealed at rest. Never shown in the studio.

capability · id only · no secret

ClickHouse

org · analytics-prod

ready

Databricks

org · lakehouse-west

ready

in a workflow Editor picks analytics-prod → import runs → next step sees the rows.

03 — SHAPE IT

Formulas you already know how to use

The steps you'd reach for in a spreadsheet — lookups, fills, joins, date maths, text cleanup. Related operations are folded together, and blank arguments resolve from the columns you've selected, so you configure less and still get the full result.

·One step, several jobs — related ops fold into a single configure-and-run

·Smart parameters — blanks fill from the columns you selected

·Same step model — configure once, chain it, re-run it

Step 2 · TEXT_JOIN

columnsfirst, last

delimiterspace

outputcustomer inferred from selection

firsttext	lasttext	customertext
Ada	Lovelace	Ada Lovelace
Grace	Hopper	Grace Hopper
Alan	Turing	Alan Turing
Katherine	Johnson	Katherine Johnson

new column customer4 rows updated

04 — CHECK IT

Every step shows its result before you commit

Configure a step and the preview underneath updates with it. A bad join, a lookup that missed, a fill that hit the wrong column — you see it in the rows immediately rather than at the end of a run.

Leave a parameter blank and the step fills it in from the columns you selected, then shows you what it chose.

·Column and sheet selection — preview only what you're working on

·Inferred parameters are shown, never silently applied

Step 4 · VLOOKUP

lookupregion_code

tableregions.xlsx › Lookup

column2 inferred from selection

region_codetext	regiontext	netnumber
NA1	North America	412.00
EU2	Europe	1180.50
APC	Asia Pacific	96.75
LTM	Latin America	530.25

matched 100%unmatched keys 0

05 — KEEP IT

Steps chain into a workflow that survives the month

Each step's output is the next step's input. The chain saves with the project, so next month you swap the source file and run it — same order, same result, none of the clicking. Steps can be reordered, and everything downstream re-runs.

·Saved with the project — reopen and run again

·Reorder freely — downstream steps re-run

1IMPORT3 files → 47,318

2TEXT_JOIN+ customer

3LOWERemail

4FILLNAspend → 0

5REMOVE_DUPLICATES− 3,128 rows

6OUTPUT44,190 rows

next month Swap customers_q3.csv for customers_q4.csv and run the same chain.

06 — EXTEND IT

When nothing built in fits, write a SQL step

Drop a SQL query into the workflow when the built-in formulas don’t cover the case. It runs on the local engine like every other step — same configuration panel, same live preview, same chaining. No rebuild, no leaving the app.

·Preview in place — errors and results show before you save

·Same step model — chains with formulas, not beside them

-- Step 5 · SQL · net_margin by region
SELECT
  region,
  SUM(spend) AS revenue,
  SUM(spend - cost) AS margin,
  ROUND(
    SUM(spend - cost) * 100.0
      / NULLIF(SUM(spend), 0),
    1
  ) AS margin_pct,
  COUNT(*) AS customers
FROM working_dataset
WHERE signup >= '2024-01-01'
GROUP BY region
HAVING SUM(spend) > 0
ORDER BY margin DESC
LIMIT 20;

Illustrative. Runs as a workflow step on the local engine — same preview and chaining as formulas.

07 — RUN IT

A processing engine, bundled as one executable

The work is done by a native Rust core built on DuckDB, shipped inside the app as a single binary. There's no runtime to install, no dependency tree to resolve and no garbage-collection pause in the middle of a join.

The interface talks to it over HTTP on your own machine. Save a workflow once, then run it again on next month's file — as often as you need. How fast and how large that goes depends on the machine in front of you.

Your window

Electron · React

Stays responsive while the engine runs the work.

HTTP · 127.0.0.1 · same machine

Bundled binary

uds-engine · Rust

DuckDB · Actix · Tokio — one executable, no separate install.

wf · March wf · April wf · May

Nothing leaves the machine. Re-run the same workflow; the machine sets the ceiling.

engine connectors compare local faq download

01 / 06

The engine

Rust underneath, so row count stops being a design constraint.

The processing core is a single Rust executable built on DuckDB and Actix — no runtime to install, no dependency tree to resolve, no garbage-collection pause in the middle of a join. The interface talks to it over HTTP on your own machine.

No artificial row cap

Files are processed in full rather than truncated to a preview limit. GB-scale CSV and Excel are the design target.

Live preview while you build

Pick the columns or sheets you care about and see each step's result as you configure it, not after you run the whole thing.

Steps chain into workflows

Each step's output becomes the next step's input. Save the chain, point it at next month's export, run it again.

Engine

uds-engine · Rust

Store

DuckDB*

Row ceiling

None imposed by the app

Formats

CSV · Excel · JSON

Studio DB

SQLite · Postgres

Org warehouses

ClickHouse · Databricks

Platforms

macOS · Windows · Linux

HTTP

Actix-web

Async

Tokio

Interface

Electron · React 18 · TypeScript

* Spill to disk · configurable memory limit. Throughput and memory figures return once measured benches pass. Builds are not code-signed yet.

Connectors

Pull from the source you already use.

Studio opens local databases on your machine. Org warehouses stay behind sealed credentials so members never see the password or token.

On your machine

Point at a file or host you control. Credentials stay in the studio profile on disk — no org portal required.

SQLite

local · .db / .sqlite

PostgreSQL

local · host you control

Org warehouses · sealed

Admin registers the credential once in the portal. Editors and viewers pick the connector by name — they never see the secret.

ClickHouse

org · secrets sealed

Databricks

org · secrets sealed

Compared with

You already have a spreadsheet. You may even have a script.

Both work right up until the file gets bigger, or the person who wrote it leaves.

Spreadsheet

Python script

Unified Data Studio

How big it can go

About a million rows, and it crawls long before that

No limit, if you tune it

No row cap — GB-scale files are the design target

Next month's file

Do the whole thing again by hand

Re-run it, if nothing moved

Open the saved workflow, swap the file, run

Who can run it

Anyone

Whoever set up the environment

Anyone — it's an app, not a repository

Seeing your work

Immediate, but easy to break silently

Run, print, squint, repeat

Live preview of every step as you configure it

Extending it

Macros, and the person who understands them

Anything you can write

SQL steps in the same workflow model as formulas

Warehouses

Export, paste, hope the columns still match

You wire the client and keep the secrets

Org connectors — ClickHouse & Databricks, secrets sealed

What it costs to start

Nothing

Python, packages, versions, a bad afternoon

One download — the engine is bundled

Where the data goes

Your disk — or a browser tab, if you're not careful

Your disk

Your disk. No account required — workflow sync optional; files stay local

Where your data sits

On your machine. That's the whole policy.

There's no sign-up and no upload step to process a file — work runs on your machine. You can optionally sync workflows (and related project artifacts) when you want; your data files stay local unless you choose to pull from a warehouse connector. That matters when the file on your desk is a customer export, a payroll run or a patient list you're not allowed to paste into a browser tab.

Works with the network off

The engine is a local process. Disconnect and file workflows still run.

Files stay on your disk

Spreadsheets and exports you open stay local. Optional sync is for workflows you choose to keep — not a silent upload of every file.

Nothing phones home

The only network calls are the ones you ask for — sync, or an org warehouse connector.

Before you download

Short answers.

The questions people ask before they trust a new tool with a real file.

Do I need an account?

No. The core app runs fully local with no sign-in. Optional sync is for workflows (and related project artifacts) you choose to keep — your data files stay on disk unless you pull from an org warehouse. Sync never gates opening a spreadsheet.

Does my data leave the machine?

Not by default. File processing stays on your disk. Network calls only happen when you ask for them — optional workflow sync, or an org warehouse connector.

What file types and warehouses?

Files: CSV, Excel, JSON. Studio connectors: SQLite and Postgres on your machine. Org sealed warehouses: ClickHouse and Databricks — members pick by name; Admins hold the credential.

Is it free?

Yes during early bird. Download the desktop app at no charge while we grow the product. Unified Data Studio is not open source — you get installers, not the source tree. Pricing after early bird will be announced separately.

Why does macOS or Windows warn me on open?

Current builds are packaged but not yet code-signed. On Mac, allow it in System Settings → Privacy & Security; on Windows, choose More info → Run anyway. Signed installers are on the roadmap.

Get it

One download. Then open your worst file.

Free for early bird — same app on macOS, Windows, and Linux.

macOS

Intel and Apple Silicon

.dmg

Download for macOS

Windows

64-bit

.exeNSIS installer

Download for Windows

Linux

x64

.AppImage

Download for Linux