Hermes Agent V0.12 Just Changed Everything

本站添加: 2026-05-10

[00:00:00]New Hermes agent update is insane. The curator release just changed everything.

[00:00:05]Imagine waking up tomorrow and your AI agent is smarter than it was today. Not because you trained it, not because you fed it more data, because it cleaned itself up overnight while you slept.

[00:00:14]That's not a future thing. That just shipped. For real. And it might be the most underrated AI release of the year.

[00:00:22]Hi, I'm Ava. I help people learn AI tools and actually put them to work in real life. I've been testing a lot of AI agents this year. And there's one pattern I keep seeing. They start great, then they get worse. Not because the model is bad, because the agent forgets how to use itself. That's the exact thing Hermes agent v0.12.0 just fixed.

[00:00:41]And the way they fixed it is wild. They taught the agent to maintain itself.

[00:00:45]Stick with me because I'm going to walk you through exactly how this works, what's new, and why this update has people calling it the biggest agent release of 2026 so far.

[00:00:54]Hermes agent is an open-source AI agent built by News Research. It came out in February of 2026. Think of it as a personal AI worker. You install it on your own machine. You connect it to your messaging apps, Telegram, Discord, Slack, WhatsApp, Signal, email, and then it just runs in the background and does work for you. The thing that makes Hermes different from every other AI tool is memory. Most AI forgets the second the chat closes. Hermes remembers every project, every preference, every solution. And when it solves something hard, it can save that solution as a reusable skill. So next time you ask, it already knows what to do. It's MIT licensed, runs on your hardware, your data stays on your machine. One command to install, done.

[00:01:37]So that's the foundation. Now here's the part that just changed. Hermes agent v0.12.0 dropped on April 30th, 2026. They called it the curator release, and the name is a perfect description of what's inside.

[00:01:51]Here's the problem this release solves.

[00:01:53]Every AI agent that has memory eventually faces the same issue, library rot. You build up hundreds of skills over weeks and months. Some are gold, some are duplicates, some are broken, some are abandoned. Your agent doesn't know which to pick, so it gets slower.

[00:02:08]It pulls the wrong skill at the wrong time, and what felt like magic 6 months ago feels frustrating now. Most people deal with this by ignoring it until the agent feels broken. Then they start over. The curator fixes that without you doing anything. It's a background agent.

[00:02:22]It runs on its own schedule. By default, every 7 days. While you're not even thinking about it, it goes through every skill in your library. It grades them, finds the related ones, and consolidates them into one cleaner skill, prunes the dead ones, and writes you a report so you can see exactly what changed. You wake up to an agent that's a little sharper than it was yesterday, every week, forever. There's a safety system built in, too, so you don't lose anything important. Pinned skills can't be touched, bundled skills are protected, hub skills are protected. The curator only touches the stuff that's safe to touch. You can also peek at what it's doing with one command, Hermes curator status. It ranks every skill by usage, so you see what's actually pulling weight in your setup and what's just sitting there. That's the headline feature, but there's a lot more in this release. The self-improvement loop got upgraded. This is the part of Hermes that decides what to remember after each task. It used to be inconsistent. Now it uses a proper grading rubric. It also prefers updating skills you just used instead of always making new ones. And it inherits the right model and credentials from the parent agent, which sounds dry, but it's a major stability fix. Four new inference providers got added, GMI Cloud, Azure AI Foundry, MiniMax with OAuth login, and Tencent Token Hub. Plus LM Studio got promoted to a first-class provider, dedicated support, health checks, live model listing. So if you like running models locally, you've got a really nice setup now. Two new messaging platforms, Tencent Yuanbao is the 18th platform Hermes supports. Microsoft Teams is the 19th. So, if your work happens in Teams, your agent can now live there with you.

[00:03:58]Spotify is now bundled in. Seven Spotify tools built right into the agent. Play music, search tracks, build queues, manage playlists, switch devices.

[00:04:08]There's a setup wizard that handles the OAuth flow for you. Google Meet is also bundled. And honestly, this one is wild.

[00:04:13]Your agent can join meetings, transcribe what's said, speak in the meeting, and follow up after with notes and action items. Two creative tools got promoted from optional to bundled by default.

[00:04:23]Comfy UI version 5 for image generation and TouchDesigner MCP for shaders, post effects, audio, and geometry. So, creative pipelines are ready out of the box. And the speed jump is real. The cold start of the terminal got cut by about 57%. Hermes launches almost twice as fast as before. Now, I want to share something that helped me get my head around all of this. When I first started exploring agent tools like Hermes, I was honestly overwhelmed. There were so many options, so many ways to break stuff.

[00:04:52]That's when I created this community called AI Profit Boardroom. It has over 2,000 members all focused on learning AI together and sharing what actually works. It taught me which workflows save time versus which ones wasted. The community shares real use cases and practical implementations every day. If you're serious about using AI to improve your work and your skills, check it out.

[00:05:14]Link in the description. Okay, let me show you what this looks like in practice. The features are cool, but they only matter if they help with real work. Here's where Hermes shines after this update. Picture your average work day. Meetings you forget the details of, notes you never wrote down, follow-ups you meant to send, reports you said you'd pull, tasks scattered across five different apps. That's the chaos this agent eats for breakfast. You hook it to your calendar and your messaging apps.

[00:05:39]It joins your meetings, takes the notes for you, sends follow-ups, pulls reports on a schedule, drafts replies in your voice, and it does it across whatever app you're already using. The old version of Hermes could already do most of that, but this version makes it sustainable because the curator keeps the agent sharp instead of letting it slowly drift into chaos. Compared to other AI agents out there, Hermes is in a different category. Most popular agents are stateless. They forget the second a session ends. So you start from zero every time. Hermes carries memory across days, weeks, and even messaging platforms. Most other agents also live in someone else's cloud. Your data passes through their servers. Hermes runs on your hardware. Nothing leaves your machine unless you tell it to. Let me give you five real use cases you can try this week. One, connect Hermes to Telegram. Now you have an AI assistant in your pocket. Text it from anywhere.

[00:06:31]Ask for research, a draft, a summary, a reminder. It remembers your preferences across every message. Two, turn on the Google Meet integration. Let it join your next call. It transcribes the whole thing, then sends you a summary with action items after. You stop being the person scrambling to catch up. Three, build cron jobs in plain English. Tell Hermes to run a morning briefing every weekday at 8:00 a.m. Or back up your project files every Sunday. No code. It just runs.

[00:06:59]Four, use the bundled creative tools.

[00:07:01]Ask for an image with ComfyUI. Build a visual effect with TouchDesigner. Both are pre-installed now, so creative work is one prompt away. Five, build a personal skill library. Every time Hermes solves something tricky, save it as a skill. Then let the curator clean up that library every week. Over time you end up with a tight, high-quality set of automations that quietly handle your routine work. Now here are some tips to actually get the most out of this update. Tip one, let the curator run. Don't disable it because you're nervous. The whole point is it does the cleanup work you'll never do yourself.

[00:07:35]Tip two, pin the skills you can't afford to lose. Pinned skills are off limits to the curator. So if a workflow is perfect, lock it down.

[00:07:44]Tip three, run Hermes curator status once a week. It shows you what skills you actually use. You'll be surprised at how many you thought you needed, but don't. Tip four, use the new one-shot mode. Hermes {dash} Z, give it a prompt, and it runs without opening the full interface. Perfect for quick tasks or scripting. Tip five, run Hermes update {dash} {dash} check before upgrading. It shows you what's coming, so you don't get surprised by changes. Tip six, try the new providers. LM Studio is great for local models. GMI Cloud is fast.

[00:08:13]Azure AI Foundry works well in enterprise setups. Variety helps you find the right fit. Tip seven, connect more than one messaging platform. Start a task in Slack, continue it in Telegram, finish it in Teams. Hermes works across all of them. If you're looking to dive deeper into AI tools and actually implement them in your work, I recommend AI Profit Boardroom. Over 2,000 people learning how to use AI effectively. Everyone shares real experiences. What's working, what's not, which tools are worth your time, which ones to skip. No hype, just solid information and practical guidance from people doing the work. It's helped me stay on top of updates like this one and figure out how to actually apply them.

[00:08:52]Link in the description if you want to check it out. If you want the full process, SOPs, and 100 plus AI use cases like this one, join the AI Success Lab.

[00:09:01]Links in the comments and description.

[00:09:03]You'll get all the video notes from there, plus access to our community of 58,000 members who are crushing it with AI. So, that's the Hermes agent curator release. The big idea is simple. Your AI agent now maintains itself. It cleans its own library, it learns smarter, it launches faster, and it now lives in more apps with more model choices than before. Install the update tonight, pin the skills you can't lose, let the curator run for a week, then come back and look at the report. You'll see exactly why this update is a big deal.

[00:09:31]I'll catch you in the next one.