Nous Research Hermes Agent: Open-Source AI With Built-In Self-Improvement Loop The MIT-licensed framework launched in February 2026 and collected 22,000 GitHub stars and 242 contributors within weeks. That pace reflects a specific appetite in the open-source AI community. People want agents that remember, adapt, and get better without requiring constant manual instruction. Hermes Agent is built around what Nous Research calls a closed learning loop. When the agent solves a task, it writes a reusable Markdown skill file, stores the outcome in persistent memory, and adjusts its approach for next time. No YAML. No manual configuration. The system handles it …