User repository on GitHub
Voice-driven control layer for macOS. Jev is the fallback, not the router. Rust core, Tauri shell.
- Forks
- 1
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Tools & CLIs
jev-use (Coco Research)
A voice-driven control layer for macOS: speak a goal and the Mac does it, watching the screen through the accessibility tree with everything running locally. Jev is the fallback, not the router — deterministic matching resolves commands in about 50 ms.
- Stars
- 2
- Language
- Rust
- License
- APACHE-2.0
- Reading time
- 3 min
- Published by
- Community
- Author
- coco-research
- Last verified
- Oct 2026
- Languages
- Rust
- Capabilities
- Open sourceRuns locally
Highlights
- Pre-alpha with no release yet: no installer, notarised build or Homebrew tap, so today you clone it and build it.
- Observation is at parity with its reference on four apps: Finder 12/12, Terminal 13/13, PI-Desktop 10/10 (Electron) and Notes 8/8, identical fingerprints.
- The voice hook is done and in daily use: 0.24–0.33 s to paste a dictation and 6–15 ms to hand a command off.
- The layer boundary is deliberate: jev-ax knows nothing about models and the decision layer nothing about the screen, so the port was proven without a model.
- Execution and the CLI are not yet done (T6/T7); until then the walk runs through crates/jev-ax/examples/walk.rs and needs macOS 26 plus a Rust toolchain.
Quickstart
bash
cargo test --workspace # 77 tests
cargo run --example list_apps -p jev-ax # the apps a walk can see
cargo run --example walk -p jev-ax -- --app Finder --json # one observation, as JSONWatch out
Copy reviewed against the repo README on 2026-10-15; it was truncated inside the 'How it works' section, so the Repo Standard material and later sections were not inspected. The parity and latency figures are the project's own published measurements, and the repo is labelled pre-alpha: watching works, acting does not yet.
Questions about this entry
Short answers drawn from the linked project or page.
A voice-driven control layer for macOS: it observes the screen through the accessibility tree and drives it, with everything running locally. Speak a goal and the Mac does it — decide whether speech was a command or dictation, resolve which capability fulfils it, observe, then act.
The GitHub description states that framing explicitly, while the captured README describes a deterministic decision layer (crates/jev-voice) and a capability registry that resolves a command in about 50 ms with no model. How and when the Jev fallback fires is not detailed in the captured section, so check the repository for the current wiring.
Done: observation (walking an app's accessibility tree into an indexed element table), the parity harness, settings (jev-config, jev-config doctor) and the voice hook, which is in daily use. Not yet: execution (T6, press/write/type into an element), the CLI (T7) and the glass-pill app UI (T12).
macOS 26 with a Rust toolchain, and your terminal needs the Accessibility grant in System Settings, Privacy and Security, Accessibility, before anything can see a screen. There is no installer, notarised build or Homebrew tap yet.
cargo test --workspace runs the 77 tests; cargo run --example list_apps -p jev-ax lists observable apps; cargo run --example walk -p jev-ax -- --app Finder --json prints one observation as JSON. The parity harness needs pyobjc and can compare against the reference with scripts/parity --live.
jev-ax is a port of a working implementation, not a rewrite, and it is judged against that reference: same elements, same order, same fingerprints, with the specification pinned by hash in reference/PINNED.json. scripts/parity supports --capture, --check and --live modes.
The README says two things live in the repository: the macOS automation runtime and the canonical copy of the Repo Standard every repository in this fleet carries. This digest's captured text was truncated inside that section, so the standard's details were not available when this entry was written.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Voice-controlled computer use on a Mac
X: Speech to live slides
X: Jev for real-time speech
X: On-device audio, no LLM in the loop
More Jev resources like this
A Mac-only voice assistant: say Hey Jev or hold right Option and Jev decides every command in one call for $0.00004 per request, while Whisper transcribes locally, Fish Audio S2.1 Pro speaks replies, and an LLM wakes only for questions.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
henryklunaris/hey-jevPython
henryklunaris/hey-jevPythonAn Ollama-style local runner for System One decision models from Hugging Face: pull, serve and query them behind the same /v1/systemone API as TypeSafe's Jev, with a terminal model manager and a browser playground.
Jev Tools & CLIs#community#catalog-import#madewithjev
Community
nvkudva/ollajevPython
nvkudva/ollajevPythonA private Mac voice diary: Parakeet transcribes your voice on-device, TypeSafe Jev splits each capture into separate thoughts and files them into Notion as notes, tasks and reminders, and nothing is saved until you approve it.
Jev Tools & CLIs#community#catalog-import#madewithjev
Community
sgaabdu4/captureDart
sgaabdu4/captureDartVellum Assistant is an MIT-licensed personal AI assistant with eight memory types, hourly proactivity, and permission-gated computer use, reachable from macOS, Windows, iOS, web, voice, email, Telegram, Slack, and Twilio.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
vellum-ai/vellum-assistantTypeScript
vellum-ai/vellum-assistantTypeScriptRemote control for coding agents: give Claude Code, Codex, Cursor, OpenCode, Pi and others a task from your phone, and get a notification when the result is ready while your laptop stays closed. Self-hosting is a free MIT install; the hosted version has a free plan and paid plans from $19/mo.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
bennykok/omg.devTypeScript
bennykok/omg.devTypeScriptAn open-source, local-first model router for coding harnesses: Jevia asks Jev for a capability tier per task, applies your safety policy, and keeps recorded execution history in your project or database so later routes learn from outcomes.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
assistant-ui/jeviaRust
assistant-ui/jeviaRustFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.