01What Hermes 0.20 Is and How It Works
Hermes 0.20 is an open-source AI agent update from Noose Research, also called the Herald release. It makes your AI agent feel like a team member you can talk to in real time. The core idea is the "open line engine" β three open lines: between you and Hermes, between Hermes and your apps, and between Hermes and other agents.
Before, working with an agent was like sending letters: you type, wait, read, repeat. Now, Hermes speaks while it generates answers, and you can interrupt it like a phone call. This closes the gaps in every exchange, making the whole interaction faster and more natural.
The update includes voice, citations, speed gains, and agent-to-agent communication.
02Main Features in the Herald Release
The Herald release packs many features, but they all serve one idea: continuous open lines. Here are the key features:
- Real-time voice β Hermes speaks while still thinking, and you can interrupt it mid-sentence by just speaking.
- Wake word β Set a custom wake word, like "hey Hermes," and it listens hands-free from across the room.
- Voice everywhere β Send voice notes on WhatsApp and other messaging platforms; Hermes replies with voice automatically.
- Grounded citations β Every research claim is backed with sources, so you can verify the AI.
- Smart approval β Hermes asks before doing risky actions like deleting files.
- Artifacts β Desktop app shows everything you've created in one place, with sandboxed live previews.
- Agent-to-agent protocol β Hermes can discover, talk to, and be driven by other AI agents.
03How to Set Up Hermes 0.20
Getting the latest version is simple, whether you use the terminal or the AgentOS. Follow these steps:
- Open your terminal and type Hermes update, then press Enter.
- If you use AgentOS, go to the Manage section and click Update.
- Wait for the update to finish; there's no install wizard and no settings to dig through.
- Set your wake word (e.g., "hey Hermes") in the settings.
- Try one hand-free request, like asking the time, to test the wake word.
That's it. The update is free and open source, with hundreds of contributors. Once updated, you can start using voice, citations, and the agent-to-agent plugin.
04How to Use Hermes 0.20 in Real Life
Using Hermes 0.20 is about natural conversation and hands-free control. Here are examples from the episode:
- Hands-free requests β Say your wake word, then ask "what time is it?" and Hermes answers without any clicking or typing.
- Dictate emails β While cooking, say the wake word and dictate a client email; Hermes drafts it before you sit down.
- Voice notes on WhatsApp β Send a voice note asking Hermes to pull together notes for your next call; the answer is waiting when you sit down.
- Mid-turn corrections β While Hermes drafts a post, type "make it friendlier" and it adjusts without starting over.
- Agent-to-agent tasks β Post one message and let Hermes and other agents (like Claude, OpenClaude) reply and build on each other's answers.
Each feature removes the old gaps of waiting, copying, and restarting.
05Results and Speed Tests in the Episode
Julian tested several features and reported concrete numbers. First, the first token speed now lands in about 0.9 seconds, down from around 4.3 seconds. That change feels huge β at four seconds there's an awkward pause; under one second it feels like the agent is just there.
The telemetry gate is 54 times faster. The desktop app is quicker across the board. When paired with newer, faster models, tasks that used to take minutes now finish in seconds.
Julian also generated images through a connected tool, and they all appeared in the artifacts view. He tested the wake word and the message went straight into the chat and answered.
06Tips, Limits, and Mistakes to Avoid
To get the most out of Hermes 0.20, keep these tips in mind:
- Use mid-turn corrections β Don't stop and restart; type corrections while Hermes is working.
- Take advantage of the wake word β Put Hermes on another monitor or across the room for hands-free use.
- Verify citations β Don't just trust the AI; click through sources to check facts.
- Approve risky actions β Smart approval will ask before deleting files, so don't ignore those prompts.
- Keep your conversation recent β Improved compression preserves recent context, so don't worry about long chats losing the plot.
- Mistake to avoid β Skipping the update because it seems like a time sink; updating takes one command and saves time later.
07Who Should Use Hermes 0.20 and Why It Matters
This update is for anyone who wants to use AI agents but feels they lack technical skills. Julian emphasizes that the skill barrier is gone: if you can talk, you can run an AI agent. You don't need to write clever prompts or know commands β just speak like you would to a person on your team.
The least technical people benefit the most because talking is something they're already good at. For business owners, Hermes can help with leads, customers, and saving hours. The agent-to-agent feature matters because it turns your separate agents into a team, and you become the manager.
The open line is here, and those who pick it up first will be miles ahead.
