Skip to content

OpenHands v1.15.0 Changelog

What changed in OpenHands agent v1.15.0, released 2026-08-21 — explained in plain English.

TLDR
  • A brand new onboarding checklist helps you get set up easily.
  • The conversation view now has an overview panel and a unified commits drawer to track your work.
  • The UI (user interface) got a lot faster, especially when AI models type out answers really quickly.
  • Tons of annoying bugs were fixed, from broken PDF previews to glitchy scrolling.

Want a heads-up when OpenHands updates?

We email you the moment a new OpenHands version drops — plus plain-English release notes like these. No spam.

Choose update emails OpenHands · Vibe Mastermind Updates
Choose which update emails you want

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.

This update is all about making the app easier to use and more reliable. It matters a lot to developers and anyone using OpenHands to manage AI coding agents, because it adds helpful new ways to track your work and squashes a bunch of frustrating glitches that used to get in the way.

What's new

  • Getting started checklist: There is now a helpful checklist in the sidebar to guide you through setting up the app, which you can turn on or off in the settings.
  • Workspace paths in Files view: When you are looking at your files, the app now shows exactly where each file lives on your computer (its "path").
  • Automations catalog: You can now install a "catalog entry" that bundles together helpful scripts, making it easier to automate tasks.
  • LLM provider connections UI: There is a new screen to help you connect and manage your AI language models (LLMs) directly to your local agent server.
  • Polished automations dashboard: The area where you manage automated tasks got a visual upgrade, including a "recommended rail" (a sidebar of suggested tools) and a smoother process for adding or importing items.
  • Conversation overview and commits drawer: When chatting with the AI, you now get an overview panel to see what's happening, plus a "unified commits drawer" that lets you view all the code changes the AI has made in one place.

Improvements

  • Faster streaming text: When an AI model generates text really fast, the app now groups those text updates together into batches. This keeps the user interface from lagging or freezing up.
  • "Skills" rename: The app finished renaming "microagents" to "skills" in the frontend (the part of the app you see), making the language much easier to understand.
  • Automations interface: The way the app handles automation tools is now powered just by its manifest (a list of what the tool does), making it run more smoothly.

Fixes

  • Silent profile downgrades: Fixed an issue where the app would secretly downgrade your AI agent's profile without telling you.
  • Missing workspace folders: In grouped view, the app used to hide some of your workspace folders if there were too many pages of them. Now it always shows all of them.
  • LLM dropdown ignored: Fixed a bug where the AI model you picked from the home screen dropdown was ignored if your agent profile had a different model pinned.
  • Onboarding modal glitches: Fixed the spacing at the bottom of the welcome pop-up and stopped leftover scrollbars from messing up the screen.
  • Code testing mix-up: The system that checks the code (CI) used to think markdown documentation was actual frontend code; that is fixed.
  • HTTPS URL search bug: Cleared out old, leftover search results for web addresses that didn't match, which used to confuse the app.
  • Git remote URL slashes: Fixed a problem where a trailing slash in a Git repository web address would break how the app reads it.
  • Localhost service URLs: Updated the app to use "127.0.0.1" instead of "localhost" for some internal service links, which prevents network glitches.
  • URL fragments preserved: Fixed a bug where the app accidentally deleted the "#section" part at the end of some web links.
  • Broken AI tool calls: Fixed how the app reads certain XML code (a way of formatting data) when an AI streams tool commands, so the commands don't break.
  • Keyboard shortcut fix: The Cmd+Enter keyboard shortcut now only triggers a build when you are in "plan mode," instead of triggering at the wrong times.
  • Skills modal overflow: If an AI skill had a really long description, the buttons at the bottom of the pop-up window used to get cut off. Now they stay visible.
  • PDF preview broken: Un-sandboxed the PDF preview window so the built-in viewer actually renders your documents again.
  • Bad model saves: Stopped the app from secretly saving your AI model choices if it failed to find your profile settings.
  • Mouse drag-to-scroll: You can now use your mouse to click and drag the "recommended rails" (the suggestion sidebars) to scroll through them.
  • Git port reading error: Fixed a bug where the app mistakenly read a Git repository's network port as part of its file path.
  • Dropped connections: Kept the events socket (the live connection that updates the app) from hanging or dropping when the app refreshes data.

Breaking changes

No breaking changes in this release.

This is our plain-language summary. Read the complete, official notes on the OpenHands official changelog ↗.

Vibe Mastermind in 00d 00h 00m 00s
Call LIVE — Join Now