User repository on GitHub
Production local Cua + TypeSafe Jev fast-path runtime for Codex and Waku
- Forks
- 1
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Agents & MCP
Jev Cua — eronmmer
A local Codex and Waku plugin for app-agnostic Mac computer use: Accessibility actions with a visual fallback, approval-gated mutations, and a Jev fast path for repetitive tasks.
- Stars
- 0
- Language
- TypeScript
- Reading time
- 3 min
- Published by
- Community
- Author
- eronmmer
- Last verified
- Oct 2026
This is one of several Jev listings for the same project. The canonical entry is Jev Cua — ronadin2002.
Used inJev for Software & DevTools
- Languages
- TypeScript
- Platforms
- macOS
- Capabilities
- Runs locallySelf-hosted
Highlights
- Launch, inspect, click, type, press keys, invoke menu items, and scroll on Mac apps, with Accessibility state first and a visual fallback.
- Visual observations return opaque grids rather than coordinates to the model; the server recaptures before a click and refuses stale pixels.
- Consequential or private actions need one-shot approval bound to the exact operation, target, and current observation.
- Every mutation requires a caller-supplied postcondition and reserves its operation key in a durable ledger before dispatch, for at-most-once execution.
- A Jev fast path recommends a workflow for repetitive tasks, which is validated locally and compiled into a deterministic adapter.
Watch out
Copy reviewed against the repo README on 2026-10-15; the README is longer than the fetched excerpt, so the full tool name list was not reviewed. The release explicitly does not establish a '1000x' speedup, no licence is declared, and screenshots cannot be reliably redacted, so secrets must never be on screen.
Questions about this entry
Short answers drawn from the linked project or page.
It is a local Codex and Waku plugin for app-agnostic Mac computer use. It can launch an exact installed app, inspect Accessibility state, take in-memory screenshots, target visually through opaque grids, click, type non-secret text, press bounded keys, invoke menu items, and scroll.
For repetitive work. The optional Jev workflow recommendation proposes a path for a repetitive task, which is validated locally and compiled into a deterministic adapter that performs the action with an exact postcondition, rather than asking the model each time.
When Accessibility targeting is incomplete, visual observation asks for informed consent before returning the exact window PNG plus 64 opaque regions, and refine returns one zoomed JPEG plus 64 opaque click actions. Screenshots go to the model and cannot be reliably redacted, so never approve a window showing secrets.
Before a visual click the server recaptures both images and refuses stale pixels; every mutation requires a caller-supplied postcondition checked before approval, again after approval, and after dispatch. A pixel change alone is never proof of the intended effect; only a verified semantic postcondition counts.
A content-free durable ledger reserves each operation key before dispatch, and the README says never to replay an attempted, interrupted, timed-out, refuted, or unknown mutation under the same or a new key; you reconcile the real app state first.
No. Text entry rejects recognizable credentials and secure fields, but the README calls this best-effort and says never pass passwords, API keys, recovery codes, or OTPs; MCP arguments and approval forms may be logged by the client, and type_text requires a verified value_equals postcondition.
No, explicitly. The README states this release does not establish a '1000x' speedup and that it still requires end-to-end measurement; what it replaces is the old Accessibility-only prototype with a usable general Mac path while keeping deterministic verification and at-most-once execution.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Voice-controlled computer use on a Mac
X: Computer use without screenshots
X: Copy and paste, with a decision in between
X: Jev driving real Chrome from Grok Bot
More Jev resources like this
A local stdio MCP server from bgivenb that runs whole browser and macOS goals for your agent: your agent supplies the goal, exact values, and completion conditions, and Flick loops observe, decide, and act with TypeSafe Jev. TypeScript core with Playwright and Swift drivers.
Jev Agents & MCP#community#github#mcp
Community
bgivenb/flick-computer-useTypeScript
bgivenb/flick-computer-useTypeScriptAn open-source computer-use MCP server for Claude Code, Codex and other agents that reads the accessibility tree instead of screenshotting every step, using Jev as a fast judge and, experimentally, as image input to spot icons and game objects.
Jev Agents & MCP#community#github#mcp
Community
8itlab/reflex-cuPython
8itlab/reflex-cuPythonVellum 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.devTypeScriptA plugin-based agent harness (FastAPI + Vue 3) bundling a chat GUI, multi-provider model access, an MCP client and marketplace, skills, workflow studio, sandboxed execution and a Tauri desktop shell. It integrates TypeSafe AI's Jev for structured Choice, Score and Noul decisions.
Jev Agents & MCP#community#github#agents
Community
andyqiuqiubo/andy-harnessTypeScript
andyqiuqiubo/andy-harnessTypeScriptA judgment layer for agents that starts with Claude Code: a PreToolUse hook asks a YAML policy you own a handful of plain-English questions about every proposed tool call, and Jev returns calibrated probabilities so thresholds — not a frontier LLM — decide allow, deny or silence, for about four cents a day.
Jev Agents & MCP#community#catalog-import#madewithjev
Community
clownware/bouncerTypeScript
clownware/bouncerTypeScriptFrom 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.