Nous Research Launches Hermes Agent: Open-Source AI That Learns and Grows

Summary

Nous Research has released Hermes Agent, an open-source autonomous AI agent with persistent memory capabilities. Unlike chatbots or copilots, Hermes runs on your own server, remembers context across sessions, and builds its own skills through experience. The agent supports 40+ built-in skills and integrates with five major chat platforms (Telegram, Discord, Slack, WhatsApp, and Signal) through a unified gateway, making it accessible wherever users are.

Hermes Agent combines several advanced capabilities: persistent memory that learns user preferences and projects over time, automated skill creation that documents and makes reusable any solution it develops, and multi-platform reach through a single gateway process. The platform also supports scheduled automations with cron scheduling, full browser and web control for automation tasks, vision analysis, and image generation. Released under the MIT License with 100% self-hosted architecture, Hermes contains no telemetry or cloud lock-in.

Beyond task automation, Hermes serves as a platform for machine learning operations, enabling teams to generate training data at scale, run reinforcement learning experiments, and export trajectories for fine-tuning models. The single-command installation works on Linux, macOS, and WSL2 without prerequisites, and features interactive setup wizards for connecting to model endpoints via Nous Portal, OpenRouter, or custom endpoints. This positions Hermes as a significant contribution to the open-source AI ecosystem, addressing the growing demand for self-hosted, privacy-preserving agent infrastructure.

• MIT-licensed with zero telemetry, no cloud lock-in, and automatic installation via a single curl command
• Features automated skill creation, scheduled automations, full browser control, and multi-platform continuation of conversations

Editorial Opinion

Hermes Agent represents a meaningful shift toward decentralized, privacy-first AI infrastructure. By combining persistent memory, automated skill creation, and true multi-platform reach in a self-hosted package, Nous Research addresses a critical gap for users and organizations unwilling to depend on centralized AI platforms. The inclusion of MLOps capabilities—particularly trajectory generation and RL training integration—signals that this isn't just a consumer tool but a foundational platform for the AI development community. If execution matches the ambitious feature set, Hermes could become a reference implementation for how open-source agents should work.