Intro

I just finished upgrading my AI assistant’s core to OpenClaw 2.21, and let me tell you… this update is insane. We’re not just talking about minor bug fixes here; they literally swapped in an entirely new brain (LLM) and massively overhauled the user experience for platforms like Discord and iOS.

For all my fellow non-developers out there, I’ve summarized the release notes so you can instantly feel how much smarter and smoother your assistant is about to become. Let’s dive in! 🏃‍♂️💨

1. 🧠 A New Brain Equipped: Official Gemini 3.1 Support!

The biggest highlight is definitely the official support for Gemini 3.1 Pro (google/gemini-3.1-pro-preview)! The previous versions were great, but 3.1 feels like a whole different dimension in terms of speed and reasoning. Watching my assistant perfectly understand my messy prompts is just incredibly satisfying.

Additionally, new providers like Volcano Engine (Doubao) and BytePlus are now supported. This includes coding-specific models, and they’ve made the API key onboarding process so much smoother.

💡 Pro Tip: You can now use the channels.modelByChannel setting to assign different models to different platforms! For example, you can use a cheaper model for your Discord server and the smartest one for your private Telegram. It’s a lifesaver for saving API token costs! 👍

2. 🎮 Discord Finally Feels Like a ‘Real’ Bot

The crown jewel of this update is the massive enhancement to Discord integration.

  • 🎙️ Voice Channel (VC) Support: Just type /vc, and your bot hops right into the voice channel! They even added an auto-join configuration for real-time voice conversations.
  • 👀 Real-time Lifecycle Reactions: Before, it was frustrating not knowing if the bot was actually processing my request. Now, it uses emojis to show you its real-time status (queued, thinking, using tools, done, or error). It’s honestly pretty cute to watch.
  • 🧵 Thread-Bound Subagents: You can now spawn “subagents” that work quietly inside a specific thread without cluttering your main chat room. Perfect for delegating complex background tasks.
  • ✍️ Live Streaming Previews: Instead of waiting for the entire response to finish, you can now see the text streaming in chunks while the bot is “typing.” Zero waiting frustration!
  • 🏷️ Forum Tag Management: You can now update a forum’s available_tags directly through channel edit actions.

3. 📱 iOS & Apple Watch Users, Make Some Noise!

If you use OpenClaw on your iPhone, you’re going to love this. The app usability just got incredibly clean.

  • Easy on the Eyes UI: They stripped out all the noisy metadata and timestamps, formatting tool outputs into concise summaries. They also added a sweet “tap-to-dismiss” feature for the keyboard in the chat view.
  • Max Level Background Survival: Your connection won’t drop even if you turn off your screen and put your phone in your pocket! The background wake and reconnect logic has been heavily optimized, so your assistant stays on standby without dozing off.
  • Apple Watch Integration: The bridge for mirrored watch notifications is much smoother now, meaning you can fire off quick replies directly from your wrist seamlessly.

4. ⚙️ Other Core System Improvements (The TL;DR Version)

This part is a bit technical, but these are solid under-the-hood upgrades:

  • Security Boost: The internal gateway lock and tool-call IDs have been upgraded from SHA-1 to SHA-256! This means the core security foundation of your assistant is much stronger.
  • Subagent Optimization: The default spawn depth for background workers (subagents) is now set to 2. This makes orchestrating multiple AI tasks much more stable and efficient.
  • Telegram Streaming Simplified: The previously complicated preview settings have been unified into a single boolean (channels.telegram.streaming), making setup a breeze.
  • CLI Convenience: When giving commands via the terminal, there’s a new defaultTo routing fallback. You no longer need to explicitly type the --reply-to flag every single time.

Wrapping Up

With the 2.21 update, you can really feel how much the developers thought about making these AI assistants feel more natural and convenient in our daily lives. The sheer intelligence of Gemini 3.1 and the mobile/Discord optimizations are sweet changes you’ll feel right away.

I’m off to put my assistant through some hardcore testing with these new features! 😂 Make sure you update your systems and enjoy a much smoother AI life!

P.S. You might have guessed it, but the first draft of this post was written by my AI assistant!