Does it look at screenshots?

It does not. Swift discovers live controls and Jev selects among available actions, with typing taken from request or screen text. There are no screenshots, no generated prose, and no per-app task scripts.

What leaves my Mac per decision?

Speech is transcribed on-device, while commands, relevant screen text, and action options go to OpenRouter or TypeSafe depending on the key prefix. Keys stay in Keychain and memory, and secure text fields are excluded.

What setup does it actually need?

Use an Apple silicon Mac with macOS 14+, Xcode command-line tools, and a funded OpenRouter or TypeSafe key entered in Settings. Then grant Accessibility, Microphone, and Speech Recognition permission in System Settings.

What happens on apps it cannot drive?

They can fail because the loop depends on live Accessibility information. There is no confirmation queue, so use the stop button or say cancel task, and quote literal text for reliable typing.

How do I install and build it from source?

Clone the repo, run bash build.sh, and open 'dist/Jev Voice.app'. It needs an Apple silicon Mac on macOS 14+ and Xcode command-line tools; it builds with Swift 6.2.3 in Swift 5 language mode and uses no package manager or third-party Swift dependencies. JEV_SIGNING_IDENTITY picks a signing identity and JEV_OUTPUT_DIR changes the output directory.

What is the difference between stopping listening and cancelling a task?

Stop listening turns the mic off while an active task keeps running, and Cancel task or the stop button stops the current task and clears queued commands. Turn listening back on with the microphone button or Option-Space; commands typed into the bar and sent with Return work with the mic off.

Does voice input work in languages other than English?

No. Speech recognition currently uses English, and for literal typing you should quote the text you want inserted, since unquoted typing requests are less reliable. You can also type commands directly into the floating bar with the mic off.

CommunityJev Demos & Experiments

Jev Cua

Mac menu-bar voice controller where Jev picks the next action from observed Accessibility controls and Swift executes it in an observe-decide-execute loop.

See highlights

Open github.com

Stars: 39
Language: Swift
Published by: Community
Author: ronadin2002
Last verified: Oct 2026

ronadin2002/jev-cua

User repository on GitHub

View on GitHub

Voice and text control for macOS. One floating bar, live UI action selection with Jev, and a continuous observe–act–verify loop.

Forks: 4
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

A small floating bar for voice and text commands. Jev picks the next action from your Mac’s live controls, the app executes it, and the loop continues toward your request. Download the full demo · 42 seconds, with sound (MP4) →. Preview above plays at 2× speed. The downloadable video plays at its original speed. Get started · How it works · Contribute. Keep talking. Turn the mic on once. Speak a full request, pause, then give another. Commands queue while work is in progress.

Quoted from the project's README. Read the full README.

Tagscommunity jevbest catalog-import awesomejev

Works forJev for Classification Jev for Routing Jev for Evaluation

Used inJev for Software & DevTools

Languages: Swift

Highlights

Queues spoken or typed commands in one floating bar that stays available across apps and full-screen Spaces.
Runs an observe, Jev choice, execute, and re-observe loop with grouped action lists and result checks against the request.
Accepts OpenRouter or TypeSafe keys selected by prefix, stored in macOS Keychain with no generative planner involved.
Exposes Jev activity traces, self-test and activity-test binaries, and English voice commands with quoted literal typing.

Watch out

Experimental Accessibility controller for Apple silicon Macs; apps with missing Accessibility info can fail and completion checks can be wrong. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

Jev-Quantum

Jev-protocol random baseline that answers noul, choice, and score questions from fast pseudorandom numbers without reading the prompt.