Nous Research (@NousResearch) introduced Bot Mode for Hermes Desktop on August 17th, turning the AI agent's existing profiles into named Bots with separate roles, models, memory, skills and profile pictures.
https://x.com/NousResearch/status/2089429432612147572 According to Nous Research's launch post, users can build a specialist Bot once, reuse it across tasks and let Bots communicate with each other. Each Bot can run a different model, giving users a way to match inference cost and capability to a particular job instead of sending every task through one configuration.
The release follows the open-source strategy that shaped Nous Research. Cofounder Karan Malhotra (@karan4d) told Fortune in 2025 that Nous Research began in 2022 as volunteers who met through Discord, GitHub and X. Hermes Agent has since become Nous Research's attempt to put the lab's model and training work inside software that can execute tools, retain information and run on a user's own infrastructure.
Profiles become products
Bot Mode packages infrastructure that already existed inside Hermes Agent into a more legible desktop interface. Hermes profiles are isolated agent instances with their own configuration, skills and sessions. Users can assign a model and description to each profile, edit its identity through a SOUL file and switch the active profile for future runs.
The profile command system also supports cloning, exporting and importing those configurations. A user can copy only a profile's configuration, credentials and identity, or duplicate its memory, skills, sessions and state as well, according to the project's command reference.
Bot Mode turns those administrative objects into characters that sit inside the main desktop workflow. A software development Bot could carry coding tools and project conventions, while a research Bot could use a separate model, memory store and set of browsing skills. A lower-cost model could handle routine work while a stronger model is reserved for tasks that justify the additional inference expense.
The memory distinction matters. Hermes stores compact persistent notes about a user's environment and preferences, and injects them into later sessions. Its skills system stores reusable procedures that can be installed, written by the user or created by the agent after completing a task. Separating both by Bot limits the amount of irrelevant context carried into each job and gives specialists room to develop different operating instructions.
The coordination test
Cross-Bot communication is the consequential part of the update. A roster of specialists becomes useful when one Bot can hand work to another without forcing the user to copy prompts, files and intermediate results between conversations.
That coordination also raises the stakes for visibility and control. Hermes agents can operate terminals, edit files, browse the web and connect to external tools. A mistaken handoff can propagate an incorrect assumption, trigger additional model spending or give a task to a Bot with broader system access than necessary. The existing profile architecture provides separate configurations and sessions; Bot Mode will be judged by how clearly Hermes Desktop presents handoffs, shared context and tool activity when several Bots participate in one job.
Hermes Desktop remains a native interface over the same open-source agent runtime used by the command line and messaging gateway. The desktop documentation describes an Electron shell and React interface connected to a headless Hermes backend. That common core lets sessions, settings, memory and skills carry across surfaces rather than trapping Bot Mode inside a separate desktop-only agent implementation.
For Nous Research, the update pushes Hermes Desktop toward a workspace for managing a small roster of persistent agents. The strategic bet is that users will prefer named, durable specialists over repeatedly configuring one general-purpose assistant. Bot Mode gives that idea a consumer-facing interface; cross-Bot work will determine whether the roster functions as a coordinated system or a collection of saved presets.