jev-chat-jarvis-mac
macOS floating chat assistant providing real-time intent, risk grading, and candidate replies.
Screen-reading companion that analyzes incoming messages and drafts contextual replies locally.
This macOS tool monitors your active chat window via system text interfaces to evaluate message intent, assess risk levels, and generate tailored reply options. It is built for professionals managing high-stakes client communications who want rapid, local intelligence without injecting code into messaging apps.
What it is
It reads screen text passively, running intent classification and risk scoring through either an online API or a local model. It then displays a floating frosted-glass panel offering customized candidate responses across multiple tones, which you can insert directly into your chat input box.
Who it's for
It suits customer support agents, community managers, and power users on Apple Silicon Macs needing instant reply suggestions during live chats. It does not suit Intel Mac users, Linux or Windows operators, or those unwilling to grant macOS screen recording and accessibility permissions.
Key features
Zero-Invasion Screen Reading
Reads chat text directly via macOS system interfaces without injecting code or hooking databases.
Intent and Risk Scoring
Classifies message intent and provides a 0-9 risk grading with actionable advice.
Concurrent Tone Generation
Builds candidate replies across 14 different tones simultaneously before local probability sorting.
Manual Region Calibration
Allows custom selection of message and input areas to prevent missed chat snippets.
Up and running in 3 steps
Our honest verdict
What's great
- Operates strictly read-only without tampering with target chat applications
- Supports local-first execution via local models or custom API keys
- Includes customizable candidate counts and tone presets like professional or casual
- Provides optional YOLO bounding boxes to highlight OCR-detected messages on screen
Where it falls short
- Requires an Apple Silicon Mac running macOS 13 or later with zero Intel support
- Demands manual granting of macOS screen recording and accessibility permissions
- Running local models demands substantial hardware resources and a ~3.8 GB model download
- Manual calibration may be required if chat layouts or contact lists shift unexpectedly
Pricing & license
Free under its MIT license. The only cost is running it yourself, which may include API usage fees if you choose to connect external cloud LLM providers rather than running a local model.
Open remote AI-training roles
874 open remote AI-training roles, updated November 15, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
Does this tool support Intel Macs?
No, it strictly requires Apple Silicon (M series) Mac computers due to local model and torch dependencies.
How does it capture chat text without injecting code?
It reads text directly through macOS system interfaces and assistive APIs rather than modifying or hooking chat databases.
Can I use my own custom LLM endpoint?
Yes, you can configure OpenAI-compatible, Anthropic-compatible, or custom API endpoints via the visual settings window.
