01What Is Hermes Agent V0.19 and How It Works

Hermes Agent V0.19 is the latest update to the open-source AI agent from Nous Research. It lives on your computer and can talk to any AI model you point it at. The nickname is the Quicksilver release because the team rebuilt how fast it wakes up, answers, and runs multiple agents.

The core idea is simple: before, you waited 4 to 6 seconds staring at a blank screen. Now, replies land in about one second. The team also added background job recovery and a gateway that routes different channels to different agent profiles.

This means you get speed, safety, and flexibility all in one update.

02Main Features: Speed, Recovery, and Multi-Agent

The release focuses on one main thing: speed. But it comes with several key features that change how you use the agent.

  • 80% faster replies – Cold start dropped from 4.3 seconds to 0.9 seconds. You see thinking notes while it loads.
  • Background job recovery – Jobs and final replies are written to a database. If the app or gateway crashes, the agent finishes and delivers when it comes back.
  • Multi-agent gateway – One gateway routes different channels to different agent profiles. Each profile can have its own model, memory, skills, and secrets.
  • Smart approval – A second AI independently assesses flagged commands. You can also set deny rules that work even in full auto mode.
  • New free models – HY3 and Laguna S 2.1 are two new free models you can plug into Hermes Agent.

03How to Set Up and Update to V0.19

Updating to Hermes Agent V0.19 is simple. You do not need to spend a weekend learning new stuff. Follow these steps:

  1. Go to your AgentOS link in the description or visit the website.
  2. Open the Manage section.
  3. Click Update Hermes right there.
  4. Wait for the update to finish – the release shipped 2,245 commits, so it might take a moment.
  5. Restart Hermes Agent if needed, and you're ready.

Once updated, you can start using all the new features immediately. The update does not change what the agent can do; it just makes everything faster and more reliable.

04How to Use V0.19: Examples and Workflows

Using Hermes Agent V0.19 is straightforward. You type a message anywhere – the desktop app, Telegram, Discord, or Matrix. The gateway checks who should answer and wakes up in under a second.

For example, you can set up multiple agent profiles on one gateway. One profile for SEO, one for writing, one for ops. The main Hermes agent can delegate tasks to sub-agents automatically using the delegate tool.

You do not need to babysit every command because the smart approval handles flagged actions.

You can also create custom workflows like the Hermes Apollo system. It uses your builds to create web pages with your voice, operate your computer, and save everything in your workspace. Hermes Oracle and Hermes Astros can monitor competitors and come up with content ideas automatically.

05Results and Comparisons: Old vs. New

In this episode, Julian compared the old version (V0.18) with the new V0.19. The differences are big.

  • Speed – Old: 4.3-second cold start. New: 0.9-second cold start. That's an 80% drop.
  • Background jobs – Old: jobs broke silently if the app crashed. New: jobs finish and deliver after a crash.
  • Multi-agent – Old: one agent answered every channel with the same brain. New: one gateway routes each channel to its own specialist profile.
  • Approvals – Old: you approved risky commands by hand, or used full auto with no breaks. New: a second AI assesses flagged commands, and deny rules hold even in auto.
  • Secrets – Old: API keys in plain text config files. New: real secrets management and a redact flag that scrubs secrets on export.

Julian also noted that the release shipped 2,245 commits in one update, showing the team is moving fast.

06Tips, Limits, and Mistakes to Avoid

Here are some tips and warnings from the episode.

  • Use the panic button – If your config breaks, boot Hermes in safe mode with all customizations off to fix it.
  • Set deny rules – Even in full auto mode, your deny rules still hold. This gives you speed with control.
  • Delegate to different models – If you get rate limited on one free model, use multiple profiles with different models (HY3, Laguna S 2.1, or a local model) and let the main agent route tasks.
  • Export sessions safely – Use the redact flag to scrub secrets before sending a session to anyone.
  • Learn now, not later – Julian says agents won't stabilize soon. The people learning while it moves will be ahead when it settles.

One limit: the update focuses on speed, not new capabilities. So don't expect dramatic new features—just a much faster agent.

07Who Should Use Hermes Agent V0.19

Hermes Agent V0.19 is for anyone who uses AI agents for tasks like content creation, SEO, or monitoring competitors. Because it's open source and works with free models, it's accessible even if you're not technical.

If you run multiple tasks in the background, the multi-agent gateway is a huge time-saver. You can have separate profiles for different jobs and let the main agent delegate automatically. This is useful for solo creators, small teams, and businesses that want to scale their output without hiring more people.

Julian showed examples from his AI Profit Boardroom community, where members manage Kanban boards, create SEO content, and build custom workflows with Hermes Agent. Even non-technical users can learn it, especially with the coaching calls and community support available.

08Smaller Upgrades You Might Have Missed

Besides the big three, V0.19 includes several smaller upgrades that improve the user experience.

  • Reasoning dials – You can adjust how much the agent thinks before answering.
  • Terminal commands – New commands for power users, like `/subscription` for billing.
  • Stacked skills – Combine multiple skills in one agent profile.
  • Real secrets management – API keys are stored securely, not in plain text.
  • Session sharing – Export sessions to Markdown or HTML, with the redact flag to remove secrets.
  • Nicer buttons – If you use Telegram, Discord, or Matrix, you get better interactive buttons.
  • Faster memory retrieval – If you use a memory system, it pulls up personalized context faster now.

These might seem small, but they make the agent easier and safer to use daily.