Does Jarvis Summarize a Paper While You Read?

Section by section, Jarvis (getjarvis.eu), a screen-aware desktop AI assistant, summarizes the paper you are reading. Press Cmd+/ on macOS or Ctrl+/ on Windows over a PDF or an arXiv page and ask for the method, the main finding, or one section. The result can be saved to Notion or Google Drive as you go. It routes the task to frontier models from Anthropic, OpenAI, or Google depending on length and reasoning needs, and can drop the summary straight into Notion, Obsidian, or Google Docs through its 30+ connectors. Your screenshots and prompts are never used to train models, tokens are encrypted with AES-256-GCM, and everything runs at rest. After a free plan (40 requests/week), plans start at $16/month.

Unlike a chatbot in a browser tab, Jarvis works on whatever is visible. When a journal PDF, a preprint on arXiv, or a paywalled abstract is on screen, you press Cmd+/ and ask. It captures the visible region, sends it to a vision-capable model like frontier models from Anthropic, OpenAI, or Google, and returns a summary scoped to that page. You can scroll to the discussion section and ask again; it re-reads the new view. There is no uploading, no pasting URLs, and no switching apps, which is what makes summarizing a long paper section by section actually fast.

A summary is only useful if you keep it. Because Jarvis ships 30+ connectors, you can say "save this as a note in Obsidian" or "add the key citations to my Notion reading list" and it writes directly through the connected app's API. It can also append to a Google Doc or create a Google Sheets row tracking papers you've reviewed. OAuth tokens for those connectors are encrypted with AES-256-GCM, so granting Notion or Google Drive access doesn't expose credentials. This closes the loop between reading on screen and building a durable literature library.

Screen-aware AI