Nous Research Ships Bot Mode for Hermes Agent

Nous Research has released Bot Mode for its MIT-licensed open-source agent, Hermes Agent, a significant update that reconfigures the agent's interaction model. Bot Mode replaces the previous system of single-agent sessions with a roster of named bots. Each of these bots represents a distinct Hermes profile, equipped with its own dedicated chat history, memory stores, skill sets, and a pinned large language model (LLM) for its operations. This architecture allows for greater specialization and isolation of tasks within the agent framework. The bots communicate with each other through a persistent "Agent Inbox," and work can be handed off between them using an "@mention" system, mimicking collaborative workflows. The feature was initially launched as a one-day public beta plugin developed by co-founder Teknium. It has now been integrated as a bundled and default-on feature within Hermes Desktop, specifically in version v0.20.3 of Hermes Agent. The deployment is immediately accessible on a desktop environment at no cost, adhering to its MIT license. This release is particularly beneficial for solo developers, startups, and small to mid-sized engineering teams who can adopt it without immediate licensing fees. However, Nous Research advises enterprises to consider it as a workstation tool rather than managed infrastructure, noting the absence of features like an admin console, single sign-on (SSO), central audit logs, and policy enforcement layers, which would be necessary for regulated environments. The intended industries for this tool include software engineering, AI research laboratories, technical media and content operations, developer relations, quantitative research, and IT operations. Potential applications are diverse, such as dedicating a research bot to a powerful reasoning model while assigning a writing bot to a more cost-effective LLM. It also enables scheduled inbox digests, nightly reports, and the creation of per-project agents that maintain strict context isolation. Furthermore, it supports complex handoff chains where different bots, like a scout, reviewer, and publisher, can pass tasks via "@mention." Internally, each bot is a Hermes profile stored under a dedicated directory, "~/.hermes/profiles/<name>/," ensuring isolated configuration, memory, skills, credentials, and chat history. This design choice leverages existing primitives within Hermes, keeping the overall complexity manageable. The creation and editing of these profiles are handled through gateway RPCs, including list, create, describe, and configure functions. Avatar generation for the bots utilizes the image.generate RPC, supporting both local and remote model interactions.