Jev Chat Jarvis
Jev is an open-source assistant for Android, Windows and macOS that reads on-screen chats, rates intent and risk, and drafts replies for you to send.
No visits from OpenBeat in the last 30 days yet. Every click on “Visit” is counted here.
Pricing and API checked by an OpenBeat editor on Nov 10, 2026 · facts from the maker’s own site ·
Tool insights
Jev reads a conversation on your screen, rates the other person's intent and the risk level, and writes three ranked reply options. It puts your pick into the input box, and you press send yourself.
Replying to a message with unclear tone or a sensitive topic can take time and cause mistakes. Jev gives a second read on the conversation and ready-to-edit wording.
Individuals who want help wording replies or reading the tone of a chat. It is a poor fit for iPhone users, people who want a fully hosted service with no API setup, or anyone who needs a verdict they can rely on without checking.
What it does
Jev (listed as Jev 聊天助手) is an open-source chat assistant from the maker Finderchangchang. On Android it reads the chat visible on your screen through the accessibility service. It then judges the other person's intent and a risk level, and writes reply drafts.
The tool never sends a message by itself. It fills the chosen reply into the input box, and you decide whether to send it.
Key features
- Reads on-screen chat text through the Android accessibility service.
- Rates intent and risk on a scale of 0 to 9.
- Drafts three candidate replies and ranks them.
- Fills the reply into the input box without sending.
- Falls back to offline Chinese OCR (ML Kit) when text cannot be read directly.
- Keeps a local knowledge base and contact profiles.
- Lets you set separate API endpoints for judgment, reply and vision.
The facts list QQ, X (Twitter) and Feishu / Lark (through the OCR fallback) as supported chat contexts.
Pricing
As of November 2026, Jev is open source under the MIT license and has a free plan. No paid plans are stated on the maker's site. You need your own API key, and the model provider bills you for usage. Those costs are not stated.
API and integrations
A public API for Jev itself is not stated, and no SDKs are listed. Jev connects to model providers instead. By default it uses DeepSeek through OpenRouter. It also works with OpenAI-compatible endpoints and Qwen (通义) compatible ones. The integration list also names 博查 Jev and TypeSafe, with no further detail.
Platforms: Android, Windows and macOS. Recent releases are Android v1.7, Windows v0.1.11 and macOS v0.6.0. The GitHub data shows the latest release tagged v1.7 on 2026-10-05.
Good to know
- Android needs version 11 or later on arm64-v8a devices.
- There is no iOS app. A web version is planned but not available.
- Judgments are probabilistic and meant for reference only.
- The model is mainly trained in English, so quality in Chinese varies.
- Group chats are analysed as one-to-one chats, which is less accurate.
- Documentation and pricing links are not given on the maker's site.
FAQ
Is Jev free to use?
Yes. As of November 2026, Jev is open source under the MIT license and has a free plan. You still need your own API key, and the model provider charges for usage.
Does Jev send messages automatically?
No. Jev fills your chosen reply into the chat input box, and you send it yourself. It drafts three ranked options to choose from.
Which platforms does Jev support?
Jev has builds for Android, Windows and macOS. The Android version needs Android 11 or later on an arm64-v8a device. There is no iOS support, and a web version is planned but not available.
Which AI models does Jev use?
By default Jev uses DeepSeek through OpenRouter. You can also set your own endpoints for judgment, reply and vision, including OpenAI-compatible and Qwen-compatible ones.
How accurate is Jev's intent and risk rating?
The maker describes the judgments as probabilistic and for reference only. The model is mainly trained in English, so Chinese results vary, and group chats are treated as one-to-one, which is less accurate.
Releases
No tracked releases yet.
In the news
No coverage yet.