Agentic AI Learning Portal

Appearance

🤖 hermes-agent ​

Nous Research's Hermes Agent is an open-source, self-improving local agent with persistent memory and a markdown-based skill system. See M12: Self-Improving Skill Generation for the full install walkthrough and a worked skill-creation example.

📊 Tool Datasheet ​

MetricDetails
Tool NameHermes Agent
CategorySelf-improving local agent, multi-platform gateway
PurposePersistent-memory agent that writes reusable markdown skills for itself and runs from a CLI or connected messaging platforms.
MemoryFile-based: MEMORY.md/USER.md (frozen per-session context) + SQLite with FTS5 for full session history search, no vector database.
IntegrationsNous Portal (300+ models via one subscription), OpenRouter, OpenAI-compatible endpoints

🛠️ Capabilities ​

  1. Procedural Skill Memory: Writes reusable SKILL.md markdown instructions to ~/.hermes/skills/ via its skill_manage tool when it solves a non-trivial, repeatable task.
  2. Frozen-Snapshot Context: Loads MEMORY.md/USER.md once per session into the system prompt for prefix-cache efficiency, rather than re-reading them on every turn.
  3. Full-Text Session Search: session_search retrieves exact past messages via SQLite FTS5 , keyword match, not semantic/embedding search.

Install: curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash then hermes setup --portal. Full docs: hermes-agent.nousresearch.com/docs.