Video: "Hermes Agent OS Just Changed Everything!" by Julian Goldie on YouTube.
What Bot Mode is
Before v0.20.3, Hermes Desktop used profiles — saved configurations you could switch between, each with its own model and system prompt. Bot Mode replaces the profile concept with something more active. Each Bot is a persistent named agent that holds its own model selection, memory, and skills separately from every other Bot on the system. Rather than switching to a profile and running a task yourself, you can have Bots message each other directly — one Bot hands off work to another and both operate in sequence without human input between them.
The collaboration rooms extend this to groups. Two to six Bots can share a room where they exchange messages, divide tasks, and resolve disagreements through a structured conversation loop. A room with a research Bot, a writing Bot, and a quality-check Bot can process a content brief end to end — the research Bot pulls Oracle data, the writing Bot drafts from that, and the quality-check Bot reviews the output and flags any section that does not meet the standard you defined in its system prompt.
The desktop setup: no terminal, 68 tools out of the box
The visual desktop interface was already in Hermes, but v0.20.3 tightens the experience considerably. Installing Hermes now requires dragging the desktop app into your Applications folder — no command-line setup. The model selector lets you choose between OpenRouter, Anthropic, OpenAI, and local models from a dropdown rather than editing a configuration file. Chat history is saved automatically, and profiles (now Bots) are managed from a sidebar panel.
The 68 built-in tools cover file operations, web search, calendar, email, code execution, and MCP connections to external services. You do not need to write any configuration to access them — they are available to every Bot you create from the point of installation.
What Bot-to-Bot messaging actually looks like in practice
Julian Goldie demonstrates a three-Bot content workflow in the video. The first Bot identifies topics from Oracle's news cache, formats them as a brief, and sends the brief to the second Bot. The second Bot drafts an article from the brief and sends the draft to the third Bot for review. The third Bot returns either an approved version or a list of specific changes, which loops back to the second Bot for revision. This runs unattended on a schedule — the only human involvement is setting up the Bots and reviewing the finished article.
The important caveat is that Bot-to-Bot messaging works inside the desktop application. It is not a distributed system where Bots run on separate machines or hand work off across a network. All three Bots in the example above are running locally on the same desktop install.
Who this update is for
Bot Mode is most useful if you are already running Hermes and want to split a complex workflow across agents rather than using a single all-purpose agent for everything. The dedicated model per Bot matters: you might use a fast, cheap model for the research and review steps and a higher-quality model for the writing step, reducing cost without sacrificing the output quality on the part that matters most.
If you are not yet running Hermes, the v0.20.3 desktop makes the initial setup considerably more accessible than previous versions. The no-terminal installation removes the barrier that previously stopped non-technical users from getting started.
Where this connects to NordSys
Configuring multi-Bot workflows — deciding which model each Bot should use, writing the system prompts, connecting Oracle and MCP tools, and testing the collaboration rooms — is the part that takes time and domain knowledge. Our AI Agents service covers setup for businesses that want these workflows running reliably without needing to figure out the configuration themselves.
See our AI Agents service →