Hermes Agent

Hermes Agent is an open-source, MIT-licensed autonomous AI agent built by Nous Research, launched February 2026 and one of the fastest-growing open-source projects of the year — crossing 230K+ GitHub stars within months. Its defining differentiator is a closed learning loop: rather than treating every session as a blank slate, Hermes builds and refines reusable "skills" from its own experience using DSPy + GEPA for evolutionary self-improvement of prompts, skills, and code, backed by a three-layer memory system (session context, persistent memory, and self-evolution). A community Skills Hub has grown to 90,000+ skills across a dozen registries, letting users install capabilities others have already built and refined rather than starting from scratch.

Hermes runs locally or on a user's own server (a small VPS is often enough for always-on use), or as a fully managed service through Hermes Cloud — a subscription with a free tier plus Plus/Super/Ultra plans ($20-$200/month) that bundle monthly usage credits, access to 200+ models, and hosted tool execution for anyone who'd rather skip server management. Either way, it's reachable through a single gateway process across six messaging integrations — Telegram, Discord, Slack, WhatsApp, Signal, and a CLI. It supports multiple LLM providers and terminal backends, so the underlying model can range from a budget open-weight model to a frontier one depending on the task.

Hermes sits in the same general-purpose Autonomous Agents category as OpenClaw rather than Development Tools → AI Coding Agents: its skill-learning loop is deliberately generic, applicable to everyday tasks and workflows rather than being scoped to software engineering inside a codebase, even though it's technically capable of coding tasks like any sufficiently general agent.