Build Emma is a desktop app for writing code. It has everything you'd expect from an
editor you use all day — files, tabs, search, a terminal, git, a debugger, language
support — and at the heart of it is EMMA, an agent you can hand real work to.
She reads your project, makes the change, runs your tests, and fixes what fails. Then she
stops and shows you the diff. Nothing touches your code until you've read it and said yes.
local-first · no account · no telemetry · works offline · cloud models optional
This repo is just the downloads. There's no source code here. It's where the
installers live, and where Build Emma looks to see if there's an update.
Download
Get the latest release →
| |
|---|
| Windows | Build.Emma_…_x64-setup.exe — the installer, and the one that updates itself · or the .msi |
| Linux | Build.Emma_…_amd64.deb (Debian, Ubuntu) / .rpm (Fedora) — installs it into your app menu, ready to pin · or Build.Emma_…_amd64.AppImage — portable: mark it executable and run it. It runs from wherever the file is and doesn't add itself to your app menu |
| macOS | Build.Emma_…_aarch64.dmg for Macs with Apple Silicon (M1 and newer) · Build.Emma_…_x64.dmg for Macs with an Intel processor |
On Windows, the installer isn't code-signed yet, so the first time you run it Windows
says "Windows protected your PC". Click More info → Run anyway. Unsigned means no
certificate has been bought — not that nothing was checked.
On a Mac, open the .dmg and drag Build Emma into Applications. It isn't signed with an
Apple developer certificate yet, so the first time you open it, macOS says it can't check it.
Click Done, then go to System Settings → Privacy & Security, scroll down and click
Open Anyway. (On macOS 14 or older, right-click the app and choose Open.) If macOS
says the app is damaged, run xattr -cr "/Applications/Build Emma.app" in Terminal and open
it again. The Mac version is new, so if anything doesn't work, tell us on Discord.
You'll also need Ollama, and the biggest model your computer can run.
EMMA's work is only as good as the model doing it — a bigger model makes fewer mistakes and
handles bigger jobs.
| Graphics memory | Model |
|---|
| 24 GB or more | ollama pull qwen3-coder:30b or ollama pull qwen3:32b |
| 12–16 GB | ollama pull qwen3:14b |
| 6–8 GB | ollama pull qwen3:8b — the smallest we'd recommend |
On a Mac with Apple Silicon, read those sizes as your Mac's memory. Intel Macs run models on
the processor, which is slow, so a Claude or OpenAI key helps most there.
For the best results of all, add a Claude or OpenAI key — bring your own, and it's kept in an
encrypted vault on your machine.
Try it, then buy it once
Free for 30 days, with everything included. If it earns a place on your machine, it's
£89, once — buy at buildemma.com.
- No subscription, no account. Your key arrives by email a minute after paying.
- It's checked on your own computer. Nothing phones home, and nothing stops working if
we ever do.
- Every 2.x update is included, on up to two of your own computers.
- 30-day refunds, no questions asked.
If the trial ends and you haven't bought, EMMA stops making changes — but your files, the
editor, the terminal and git all keep working, and nothing is locked or deleted.
The full terms are short and in plain English: licence ·
privacy.
Updates
Build Emma checks this page for a newer version when it starts, tells you when there is
one, and installs it only when you say so. Every update is signed, and your copy refuses
one that isn't. You can turn the check off in Settings → Updates.
Checking a download
Every release lists a SHA-256 checksum for each file, and includes them in SHA256SUMS:
sha256sum -c SHA256SUMS --ignore-missing
Help
