v1.5.9.1a · 05 Jul 2026 Latest

One prompt.
One compiled desktop app.

From an idea to a shareable installer — entirely on your machine.

Operator X02 turns engineering ideas into finished native tools — not just generated code.

Describe the app you want. Operator X02 scaffolds it, generates the code, compiles it with cargo and Tauri, and hands you a real Windows .exe and installer — without leaving the editor, and without touching a cloud.

commit 979deb4  ·  Windows 10 / 11 (64-bit)  ·  MIT licence  ·  no account

How Build Mode works

A sentence goes in. An installer comes out.

No terminal gymnastics, no packaging config, no per-OS build scripts.
Build Mode runs the whole compile-and-bundle pipeline for you.

You type
Creating project filessrc-tauri/ · package.json
Generating code4 files written
Installing dependenciesnpm install · cargo fetch
Compiling & bundlingcargo build --release
You get
Windows Ready
.msi .exe
Linux In progress
.deb .rpm .AppImage
macOS Soon
.dmg .app

Windows installers ship today. The Linux pipeline already packages — see platform status.

Other AI tools

Hand you source code, or a web app hosted in someone else's cloud. You still open a terminal and figure out packaging for every OS yourself.

Operator X02

Hands you a file a user can double-click to install — Windows and Linux from the same prompt, compiled locally, offline.

Caught in the act

One sentence. Sixteen files. One installer.

This is Build Mode compiling NoteApp — a real Tauri + React desktop app — from a single prompt. Not a mockup of the IDE. The actual IDE.

Building · 1m 39s Operator X02 IDE running Build Mode: the NoteApp project tree with 16 generated files on the left, App.css open in the editor, and the build panel compiling the Rust release build.
Operator X02 · Build Mode generating NoteApp · 16 / 16 files written · compiling the Rust release build locally

You typed one line. Build Mode wrote a whole project: a React front end, a Tauri Rust shell, and every config file that ties them together — then started compiling it into something you can install.

Generated — NoteApp/
src/the React app · 4 files
└─index.htmlloads main.tsx
  └─main.tsxrenders App.tsx
    └─App.tsximports App.css
      └─App.cssthe file open ↖
src-tauri/native shell · 3 files
└─tauri.conf.jsonwindow + bundle
  └─main.rswraps the web build
vite.config.tsbuild config
package.jsondependency manifest

index.htmlmain.tsxApp.tsxApp.css
wrapped by src-tauri/ → compiled to a native binary

16
source files
written & wired
~1–3 GB
dependencies
pulled & compiled locally
~5 MB
installer
the thing you ship

The gigabytes are the build, not the app — the Rust toolchain, npm packages and cargo crates compiled on your machine. That work is why a cold build takes the 30–45 minutes you see, and why the installer that comes out is only a few megabytes. Figures are typical for a Tauri release build and vary with the project and machine.

New in this release

Build Mode — the whole workflow, one flow.

Not a code generator with extra steps. The path from a sentence to something a colleague can install runs end to end, on your machine, in the editor you already have open.

  1. 01Ideaa tool the bench needs today
  2. 02Promptone line, plain English
  3. 03Projectscaffold + Tauri config
  4. 04Codemulti-file, backed up first
  5. 05Compilecargo · release build
  6. 06Installer.exe + .msi, ready to hand over

Six steps you used to run by hand. Now one.

Inside step 03 – 06

What the IDE does while you wait.

SCAFFOLD
Project structure

The target is detected from your sentence, the project tree is created, and the Tauri config is written for you.

GENERATE
AI writes the code

Multi-file generation through the same surgical edit pipeline the rest of the IDE uses. Every write is backed up first.

COMPILE
cargo · tauri

A real release build. Cold Rust compiles are slow, so the panel shows a moving bar and a live clock rather than a frozen 100%.

PACKAGE
.exe + installer

Out comes a signed-by-nobody, double-clickable Windows binary and an installer you can hand to someone who has no toolchain.

Prompt in. Installer out. Nothing left the machine.

Why it matters

Generating code is the easy part.

The gap between a good answer and a running application is six manual steps long. Build Mode closes it. No competitor named — this is a comparison of workflows, not of products.

Generate code, then do the rest yourself
01 Ask for the code
02 Copy it out of the chat manual
03 Create the project manual
04 Fix the dependencies manual
05 Get it to compile manual
06 Package an installer manual
You still have a folder, not an application
Build Mode
01 Describe the application
A ready-to-install executable
Workflow capabilities compared
CapabilityCode-generation workflowBuild Mode
Generate the code
Create the projectPartial
Resolve the build pipelineManual
Compile an executableManual
Package an installerManual
Runs with no network with a local model

Build Mode itself never calls out. Code generation uses whichever AI provider you configured — point it at a local model and the whole pipeline runs offline.

Pick your target

Four routes, no guessing.

A build request now opens a card with four routes instead of the IDE deciding for you. Choose one to see what it does.

Or skip the menu

› build a CAN log viewer for windows

Say it plainly and Build Mode goes straight to a native build. The word windows picks the desktop route; the app name — can-log-viewer — is detected from the sentence. No card, no configuration.

Why desktop

Because a browser cannot open a COM port.

Most AI tools generate web apps. Engineering tools need to touch the machine they run on.

Serial / COMUSBCANSPI · I²CCameraGPU / CUDAVendor driversIndustrial PCsAir-gapped networks

Engineering tools need direct access to hardware — and often to no network at all. That is why Build Mode compiles native desktop applications, not pages that ask a browser for permission.

What people build with it

Not just demos.

Automotive
CAN bus analyser
calibration utilitytest-report generator
→ installer
Robotics
Camera calibration tool
actuator monitorteleop dashboard
→ installer
Manufacturing
PLC configurator
line-monitoring dashboardfixture test rig
→ installer
Edge AI
Jetson dashboard
inference monitortegrastats viewer
→ installer
Test & lab
Sensor monitor
offline diagnostic utilitydata-capture tool
→ installer
The shapes it was designed around

Five tools, one afternoon each.

Serial monitor

Live-plots a COM port and logs to CSV. No Python environment to set up on the bench machine.

Hardware access
CAN log viewer

Parses .blf and .asc logs offline. Vehicle data never leaves the laptop.

Offline
Jetson dashboard

A small native binary with live gauges, light enough to run on the edge device it is watching.

Edge hardware
Calibration utility

An installer a technician double-clicks. No toolchain, no Node, no README on the production floor.

Hand-off
Test report tool

CSV in, charts out, limits checked, PDF signed off. Fully offline, which is usually the requirement.

Offline

Whatever the bench needs on Tuesday.
Described in one line.

A tool you can hand to someone who has no toolchain.

Platform status

Windows today. Linux in progress. macOS next.

The same Build Mode, the same prompt — three sets of native installers. Here's exactly where each one stands.

Ready · Shipping
Windows
10 / 11 · 64-bit
OperatorX02.msi
OperatorX02.exe

The main product, in daily use. Download, run, and build native apps from a prompt right now.

In progress
Linux
Ubuntu / Debian / Fedora
OperatorX02_amd64.deb
OperatorX02.x86_64.rpm
OperatorX02.AppImage
Compiles & packagesGUI validation next

Backend builds clean; .deb + .rpm already package. Verifying the desktop window on real Linux before we call it ready.

Coming soon
macOS
Apple Silicon & Intel
OperatorX02.dmg
OperatorX02.app

On the roadmap. The cross-platform core is already in place — macOS bundling follows once Linux ships.

v1.5.9.1a ships Windows installers only. Linux and macOS builds land in their own releases — watch the repo.

Installation

Download an installer.

No build required. Every feature is included — there is no paid tier and no account.

Download the .exe or .msi from GitHub. Open the release page above (or use the Download button) to grab the installer you need. Windows SmartScreen may appear. Click More infoRun anyway. This is expected for unsigned indie apps — the binary is built from the MIT-licensed source in this repository, and you are welcome to build it yourself instead.

Where this fits

One release in a longer argument.

Operator X02 is not accumulating features. Each release adds one thing an engineering IDE needs and a general-purpose one does not — and two more are already taking shape.

Shipped
Engineering Memory

The AI remembers your architecture, your hardware and the reason behind every decision. Read more →

This release
Build Mode

It can now turn what it knows into a compiled, installable application — without leaving the editor.

In progress
Something for the bench

In active development. If you have ever fought a terminal to talk to a board, you will want this one.

Where it goes
The bigger idea

Memory and Build Mode were the first two moves. What they add up to is still under wraps — for now.

The last two are direction, not a dated commitment. The roadmap lives on GitHub.

— feel it. enjoy it. —

Stop generating code.
Start generating software applications.

Describe the tool. Ship the application. Free forever, MIT licensed — no account, no telemetry, no cloud lock-in.

MIT License  ·  No account  ·  Runs offline  ·  Windows 10 / 11