Video: "Hermes Agent OS 3.0 Just Dropped..." by Julian Goldie on YouTube.
The Obsidian memory layer
The biggest practical change in 3.0 is how memory works. Previously Hermes stored context in its own internal format — useful but opaque, and not connected to anything else in your workflow. The Obsidian integration changes that by linking the agent's memory directly to an Obsidian vault you control.
What that means in practice: everything the agent learns about your business, your preferences, and your previous work gets written into readable, searchable notes. You can review what Hermes knows, edit it, or feed it specific files before a job. That two-way connection makes the memory layer genuinely useful rather than a black box. It also means if you switch agents or tools in the future, your context isn't trapped in a proprietary format.
Grok integration — adding a visual layer
Earlier versions of Hermes were strong on text but had no visual capability. Plugging in xAI's Grok adds image generation, X (Twitter) search, and basic visual analysis to the same agent workflow you already use for text. That combination — research, text, images all in one pipeline — removes the step where you have to switch tools to generate a supporting asset for a piece of content.
In practice, the image quality from Grok is appropriate for internal documents and basic social content. It is not a substitute for a proper design workflow on anything client-facing. But for the kind of quick-turn content a small business produces regularly, it is a useful addition.
ChatGPT data syncing
The 3.0 update includes the ability to pull conversation history and context from ChatGPT into the Hermes memory layer. If you have been using ChatGPT for customer research, product development, or content planning, that accumulated context can now live inside Hermes's shared memory rather than sitting dormant in a separate chat history.
This is useful if you are migrating towards Hermes as your primary agent but have months of useful context in ChatGPT. Worth knowing: the import is not automatic — it requires a specific setup step.
The redesigned mission control dashboard
The dashboard in 3.0 reorganises the control surfaces significantly. Agent status, skill management, Kanban board, memory state, and API health are all accessible from one panel. Earlier versions required navigating between separate views or using the terminal for some of these. The unified layout matters most for teams running multiple agents simultaneously — being able to see the full picture without switching tabs reduces the cognitive overhead of managing a swarm.
Where this connects to NordSys
We set up and maintain Hermes Agent OS installations, including 3.0 upgrades for businesses already running earlier versions. Whether that is configuring the Obsidian memory layer, connecting Grok, or importing legacy context from ChatGPT — the upgrade involves decisions that benefit from knowing what your existing setup looks like before you touch anything.
See our AI Agents service →