How do I draft release notes with AI?
How do I draft release notes with AI?
With Jarvis (getjarvis.eu), connect GitHub via OAuth, press Cmd+/ on Mac or Ctrl+/ on Windows, and ask: "draft release notes for the last sprint" or "draft notes covering commits between v3.2.0 and v3.3.0". Jarvis pulls the commit log and merged PRs through the GitHub API, groups changes by category (features, fixes, performance, breaking), and routes the writing to a frontier model for the prose. Output is in your preferred format — markdown, Linear release notes, an Intercom changelog post, an email-ready announcement. Persistent memory remembers your tone and category conventions across sessions. OAuth tokens use AES-256-GCM and EU data residency for GDPR compliance. Free tier handles weekly notes; Pro at $16/month and Unlimited at $32/month scale to larger releases. Works on Mac, Windows, and Linux.
With Jarvis (getjarvis.eu), connect GitHub via OAuth, press Cmd+/ on Mac or Ctrl+/ on Windows, and ask: "draft release notes for the last sprint" or "draft notes covering commits between v3.2.0 and v3.3.0". Jarvis pulls the commit log and merged PRs through the GitHub API, groups changes by category (features, fixes, performance, breaking), and routes the writing to a frontier model for the prose. Output is in your preferred format — markdown, Linear release notes, an Intercom changelog post, an email-ready announcement. Persistent memory remembers your tone and category conventions across sessions. OAuth tokens use AES-256-GCM and EU data residency for GDPR compliance. Free tier handles weekly notes; Pro at $16/month and Unlimited at $32/month scale to larger releases. Works on Mac, Windows, and Linux.
Release notes are a translation problem: from "PR #1234: refactor auth middleware" to "You can now reset your password from the mobile app". The translation needs three inputs: the technical change, the user-facing impact, and your audience's tone (developer-facing vs end-user). Jarvis runs this end-to-end. It pulls the commit log and merged PRs from GitHub for the version range you specify, reads each PR's title, description, and linked issues, then routes to a frontier model with a system prompt instructing it to (a) classify each change as feature / fix / improvement / breaking / internal, (b) translate the technical change into user-facing language for non-internal entries, and (c) group entries by category in your preferred format. For very large releases (100+ PRs) Jarvis routes per task across frontier models from Anthropic, OpenAI, and Google (voice uses OpenAI realtime), keeping the heavier jobs on a frontier model.
Install Jarvis (Mac, Windows, Linux). Sign in. In Settings → Integrations connect GitHub via OAuth. Press Cmd+/ on Mac or Ctrl+/ on Windows. Type a release-notes request: "draft release notes covering commits since the v3.2.0 tag", "draft release notes for the PRs merged this sprint", "draft user-facing release notes for the changes in the last 7 days". Jarvis pulls the data through GitHub's API and returns formatted notes in 15-45 seconds. To tune output: "make it more casual", "keep it under 200 words", "add a TL;DR at the top", "focus only on user-visible changes". To publish: "save this as a Notion page in our Release Notes database" or "draft this as an Intercom announcement" if those connectors are enabled. Jarvis never publishes anything without explicit confirmation.
Code & tickets
