Review the diff open in front of you.

A diff open in GitHub is enough for Jarvis (getjarvis.eu), a screen-aware desktop AI assistant, to review the change. Press Cmd+/ on macOS or Ctrl+/ on Windows over the pull request and ask for a summary, the risky lines, or what a reviewer would question. It works from what is rendered on screen, so you never paste the diff into another tool. Through its GitHub connector it can fetch the linked commit or issue, and it can draft review comments or post a summary to a Slack thread via its 30+ connectors. Your code screenshots are never used for model training, tokens are encrypted with AES-256-GCM, and everything runs at rest. After a free plan (40 requests/week), Jarvis is $16/month.

Code review starts with understanding the change, and Jarvis reads the diff as GitHub shows it. With the Files Changed tab on screen, ask it to explain the intent, flag risky edits, or spot a missing null check. Because it reads the visible diff rather than the whole repo, it stays focused on what the PR actually touches. Scroll to another file and ask again; it re-reads the new hunk. Routing to code-strong models like frontier models from Anthropic, OpenAI, and Google means it catches logic issues and edge cases, not just style nits.

A diff rarely tells the whole story, so Jarvis uses its GitHub connector for more. Ask it to fetch the linked issue, show the commit that introduced a touched function, or compare against the base branch, and it queries the GitHub API. It can then draft review comments for you to post, or summarize the PR for the team in a Slack channel or Linear ticket. Connectors authenticate with OAuth and AES-256-GCM-encrypted tokens, so wiring up GitHub is secure. This blends what's on screen with repo history for a more grounded review.

Screen-aware AI