Verified project record

jev-chat/jev-chat-jarvis

jev-chat-jarvis is an Android-first assistant that reads on-screen chat conversations, judges the other person's intent and the moment's risk, and drafts replies that it fills into the message box for the user to send. It works as an overlay on WhatsApp, QQ, Feishu and X without logging into or modifying the chat app.

Project overview

It analyzes intent, risk and response strategy before drafting a reply, rather than generating a reply immediately, and it never sends messages on its own.

Project type: AI Agent
Use cases: Documents & Office
Deployment: Refer to project documentation
License: MIT

Best for

General users on Android 11 or later (Chinese edition requires an ARM64 phone) who want help judging intent and risk in a chat conversation and want drafts placed into the message box while keeping manual control over sending.
Users who already have a model provider account and API key and are willing to configure it, since chat text is sent to that user-configured provider for processing.

Key capabilities

A judgment model reads the on-screen conversation and works out the other person's intent, the risk of the moment, and the best way to respond before any draft is written.
Reads the conversation via the Android accessibility service, desktop chat window screenshots with on-device OCR or system text APIs, or copied messages on iOS, without modifying or logging into the chat app.

Limitations and risks

A model provider configured by the user with their own key is required, and chat text is sent to that provider; the data boundary is external and optional only in the sense that no bundled provider is forced, but inference is not documented as local.
The tool fills the reply into the message box only; it never sends messages on its own.
Android 11 or later is required, and the Chinese edition needs an ARM64 phone. Desktop and iOS paths rely on OCR, system text APIs, or copied messages.
GPU requirements, cost dependency, and telemetry practices are not documented, so those factors cannot be assessed.

Getting started

Download jev-assistant-v1.7-release.apk, install and open the app, and follow its guide to turn on the accessibility service and the overlay permission. No coding is required. The app runs on Android 11 or later; the Chinese edition needs an ARM64 phone. You also need to configure a model provider with your own API key.

Evidence and sources

GitHub project description: The chat decision assistant: before you reply, Jev reads the chat, judges intent and risk, and drafts replies you fill in with one tap. You press send. Android · Windows · macOS ·…
README: Before you answer, Jev works out what the other person really means, how risky the moment is and how to respond, then drafts replies you can fill in with one tap.
README: Its Jev judgment model reads the conversation and works out the other person's intent, the risk in the moment and the best way to respond. Only then does it draft replies, and it…
README: | Android | [Jev Chat Assistant](../cn/README.en.md) (Chinese edition) | Overlay on QQ, Feishu, X and WhatsApp | v1.7 |
README: 1. **Read.** Jev reads the conversation on your screen: through the accessibility service on Android, the chat window on desktop (screenshots with on-device OCR, or system text AP…

Source updated at: 2026-10-06T18:24:54Z

Open original GitHub project