Nous Research Launches Open-Source Hermes AI Agent Desktop App

Nous Research’s Hermes Desktop is designed to make AI models accessible on personal computers, where local inference can keep data on-device and reduce reliance on cloud services or an internet connection. Adoption has nonetheless remained difficult for mainstream users because running open-source models typically requires selecting compatible model files, downloading large weights and configuring an inference engine for a machine’s specific hardware.

Nous Research has now added a one-click local model setup feature to Hermes Desktop. The software automatically detects a computer’s hardware, downloads suitable model weights and configures the llama.cpp inference engine, removing several manual installation steps. The feature supports macOS, Windows and Linux, allowing users to launch compatible open-source models locally without managing complex settings or determining hardware requirements themselves.

Nous Research’s open-source Hermes Agent is designed to help developers and AI research teams build agent-based workflows. Its new Bot Mode expands the system beyond a single agent conversation by organizing multiple named bots into a roster. Each bot can be assigned a distinct role and model, allowing users to structure automated work more like a team with specialized responsibilities.

Nous Research said Bot Mode gives every bot its own memory, skills and designated model, while an Agent Inbox enables bots to communicate and hand tasks to one another. The feature is included by default in Hermes Desktop v0.20.3, removing the need for a separate installation and giving developers a ready-made framework for deploying multi-agent research, software development and operational workflows.

Nous Research’s Hermes Agent is an open-source AI agent project built around a persistent operating layer and a closed-loop learning system. It organizes experience gained from completing tasks into reusable Skills and continuously refines how it works. The project is designed to move AI beyond one-off question-and-answer tools and turn it into an always-on digital workforce that can collaborate across platforms and build capabilities over time.

As of July 19, 2026, Hermes Agent had launched its /learn feature, allowing users to convert three types of content—local files, webpages and conversations—into Skills with one click. It also supports integrations across multiple platforms and inference engines for a range of AI models. Recent tutorials have focused on migrating OpenClaw memory to Hermes Agent, lowering the barrier to transferring existing personal knowledge and workflows.

NousResearch’s Hermes Agent is an open-source AI agent that can run tasks locally and connect to models and development tools. The significance of v0.14.0 lies in bridging consumer subscriptions and developer APIs. Users can reuse their web-based authorization for Claude Pro, ChatGPT Pro and SuperGrok, lowering the barrier of purchasing separate API usage. No new fees or subscription price changes were announced with the release.

As of July 20, 2026, NousResearch had released Hermes Agent v0.14.0 with a new “hermes proxy” command that converts authorization from the three AI services into a standard, OpenAI-compatible local API. The update also elevated xAI to a native core capability and added LSP semantic diagnostics, giving the agent real-time error and type information as it generates code. xAI separately released an official installation guide.