01What Is Hermes Agent V0.12?
Hermes Agent V0.12 is a free update to the open-source AI agent that lives on your computer and does tasks for you. It is 100 times more powerful than the previous version, with a built-in janitor called the curator that cleans up your skills automatically.
Your agent can now run in the background, join Google Meets, control Spotify, and improve itself without you touching anything. This release is called the Curator Release, and it was built by 213 community contributors with 1096 improvements.
02Main Features and What’s New
The biggest new feature is the autonomous curator. It runs every seven days in the background, grades your saved skills, merges similar ones, and deletes useless ones. This keeps your agent fast and smart without you doing any work.
Other major updates include:
- 57% faster startup – the agent loads only what it needs, so it is ready almost instantly.
- Spotify integration – you can play music, search songs, and manage playlists through your agent.
- Google Meet plugin – your agent can join calls, transcribe them, speak, and follow up afterwards.
- Two new chat platforms – Microsoft Teams (19th) and Tencent Yunbao (18th) are now supported.
- Four new AI providers – GMI Cloud, Azure, Minimax, and Tencent are now available.
- LM Studio support – run local models like Nemetron 3, DeepSeek v4, and Gemma 4 for free with your agent.
- Comfy UI and Touch Designer – optional add-ons for AI image/video generation and visual art.
- Remote model catalog – new models from Open Router show up without needing to update Hermes.
03How to Update to Hermes Agent V0.12
Updating to the latest version is simple. Follow these steps:
- Open your terminal on your computer.
- Type Hermes update and press Enter. The update will fetch and install automatically.
- Once done, type Hermes to start the agent. You will see it is now on the latest version.
- To run the curator, open a new terminal tab and type Hermes curator run. This starts the background cleaning process.
- Check the status anytime by typing Hermes curator status. You will see the last run, summary, and what was changed.
No other steps are needed. The curator will run every seven days on its own after the first trigger.
04How to Use the New Features: Spotify, Google Meet, and More
Spotify control: Type Hermes off Spotify to link your account. After signing in securely, you can tell your agent to play music, search songs, or change your playlist. It works through a simple command.
Google Meet integration: First, run Hermes meet install to set up the plugin. Then use Hermes meet join [URL] to have your agent join a meeting. It will enable captions, create a transcript file, and even speak on the call.
If you miss a meeting, your agent can go and give you the details later.
Using local models with LM Studio: Open LM Studio and see available models like DeepSeek v4 or Quen 3. Your Hermes agent can use them for free. This lets you run AI locally without paying for cloud services.
Command line tricks: If you type a message and want to delete it, press Ctrl + X. The agent also renders math equations (LaTeX) properly now.
05Results and Comparisons from the Episode
Julian tested the Google Meet plugin with a similar tool called OpenClaw and found it was very buggy. He says Hermes Agent is way more powerful and smoother. If he had to choose between the two, he would use Hermes now.
The curator feature means your agent gets smarter over time without you doing anything. Old skills that slow things down and waste tokens are removed automatically. This is a big improvement over the previous version where skills would pile up.
The startup time is 57% faster. Julian noticed OpenClaw has become slower when opening, but Hermes is now lightweight and ready almost instantly. The remote model catalog also means you don't need to update Hermes to get new AI models – they appear automatically.
06Tips, Limits, and Mistakes to Avoid
Don't skip the curator setup. Run Hermes curator run at least once to start the automatic cleaning cycle. Without it, old skills can pile up and slow your agent down.
Be careful with Google Meet. The plugin works well with Hermes, but Julian says it was buggy on other tools. Make sure you run Hermes meet install before trying to join a meeting.
Use LM Studio locally. It is free and works with Hermes. You can test new models like DeepSeek v4 without paying. Just open LM Studio and search for models there.
Remember the 19 messaging platforms. Your agent can be controlled from Teams, Yunbao, or any supported chat app. Use this to manage your agent from anywhere.
Don't forget to update regularly. Type Hermes update to get new features. The curator runs every seven days, but you should check for updates manually too.
07Who Is Hermes Agent V0.12 For?
Hermes Agent V0.12 is for anyone who wants a free, powerful AI agent that works on their own computer. It is great for people who want to automate tasks like joining meetings, controlling Spotify, or cleaning up their AI tools without manual work.
If you are a developer or power user, you will like the 19 messaging platforms and the ability to run local models with LM Studio. The curator saves you time by maintaining your skills automatically.
For business users, the Google Meet integration means you can have an AI assistant that transcribes and follows up on calls. And because it is open-source and free, there is no subscription cost. This update makes Hermes a serious competitor to paid tools like OpenClaw.
08Final Recap of the Curator Release
Hermes Agent V0.12, released on April 30, is called the Curator Release. The biggest new feature is the autonomous curator that cleans up your AI skills every seven days. It also starts 57% faster, works on 19 messaging platforms, and supports Spotify and Google Meet natively.
Four new AI providers were added: Azure, AI Foundry, Minimax, and Tencent. 213 community contributors helped build this release, with 1096 code changes and 360 bug fixes. To update, just type Hermes update in your terminal. To run the curator, use Hermes curator run.
Thanks to the self-improvement loop, your agent can now grade and update itself safely without touching your shell or web. This is a much more powerful autonomous AI agent that keeps getting better on its own.
