Hermes Desktop Launches Public Beta: Cross-Platform AI Agent with Cross-Session Memory & Self-Evolut

Hermes Desktop Public Beta Released – Native Windows/macOS/Linux APP Now AvAIlable

If you have followed the AI agent space for a while, you may have come across a project nicknamed "horse raising" by Chinese developers — referring to Hermes Agent. Developed by Nous Research, Hermes Agent is an open-source AI agent distinguished by its cross-session memory and self-evolution capabilities, allowing it to continuously accumulate experience and Skills through user interaction.

On June 3, Nous Research officially released Hermes Desktop, a public preview beta available for download across all major platforms.

Previously, Hermes Agent primarily ran via command-line interface (CLI) or terminal text interface (TUI). Among its user base in China, some deployed it on servers, others used third-party web UIs or desktop wrappers, and some integrated it with Feishu (Lark) to "raise their horse." The official desktop release fills this gap. For an AI Agent positioned as a "digital employee," native desktop Operation is a major leap in user experience — running directly on a user's computer without relying on a browser or rEMOte server.

Hermes Desktop currently supports macOS 12 and above (DMG installer), Windows 10/11 (EXE installer), and LinUX (via terminal installation command). Installation packages for all three platforms are now available, veRSIon v0.15.2. The desktop app shares the Same configuration system as the CLI/TUI — sessions initiated on desktop can continue in CLI or TUI, and vice versa, preserving state seamlessly. This means a user can start a task on desktop, continue it on mobile via Telegram or Discord while away, and resume in the terminal at home.

At its core, Hermes’ capabilities revolve around memory and skill accumulation. It learns the structure and context of user projects and automatically generates reusable skill modules. When solving problems, it records successful paths for future reference in similar tasks. Task planning, tool calling, observation feedback, and cross-session state retention form a closed-loop learning mechanism. For users handling repetitive workflows, Hermes can automate tasks that once required repeated configuration into reusable skill templates — genuinely becoming smarter with use.

On the tooling front, Hermes supports Web Search and browser automation, visual understanding, Image Generation, text-to-speech (TTS), and integration with external tools and services via MCP (Model Context Protocol). It also supports local runtime, Docker containers, SSH remote connections, and Singularity backends, with SECurity hardening options for containers. Its scheduling capability allows users to set up timed tasks using natural language — for example, auto-generating reports, performing backups, or sending daily newsletters — no scripting required, just everyday language.

For multi-account and multi-task scenarios, Hermes offers isolated sub-agent mechanisms — each sub-agent has its own independent conversation window and terminal environment, preventing interference. This is ideal for Professionals handling multiple projects or client scenarios simultaneously.

Hermes Agent’s open-source nature (MIT license) is a key driver of its commUnity ACTivity. Before the official desktop release, the community contributed various third-party desktop wrappers and web UIs, demonstrating strong user demand for graphical interfaces. While the official version now provides the most Standardized and stable experience, third-party solutions remain valuable — especially for custom integrations like the Feishu setup popular among Chinese users.

For the "horse raising" community, Hermes Desktop is more than just another runtime — it transforms a developer-focused tool into a native desktop application that any user can install and operate locally via a graphical interface. Paired with Nous Portal’s multi-tier subscription plans (Free, Plus, Super, Ultra), users can access AI capabilities across 300+ frontier models. This highlights another key difference between Hermes and other AI agent products: it is not tied to a single model provider, allowing users to flexibly switch between or mix models from different providers as needed.

With the desktop release, the bARRier to using Hermes drops further. For users unfamiliar with the command line, the graphical interface offers a more intuitive way to interact. For technical users, seamless cross-platform switching ensures Productivity remains unaffected. This design philosophy — combining graphical ease-of-use with command-line flexibility — represents a mainstream direction for modern AI tools.

Source: https://hermes-ai.net/zh/desktop/