NousResearch/hermes-agent
The agent that grows with you
What it solves
Hermes Agent is a self-improving AI agent designed to move beyond static LLM interactions. It solves the problem of "forgetting" and lack of specialized capability by implementing a closed learning loop that allows the agent to create, refine, and remember skills and user preferences across different sessions and platforms.
How it works
The agent operates as a flexible gateway that can be connected to various LLM providers (via Nous Portal, OpenRouter, OpenAI, etc.) and deployed across multiple environments (local, Docker, SSH, or serverless via Modal/Daytona). It uses a combination of agent-curated memory, FTS5 session search for recall, and a skills system compatible with the agentskills.io standard to autonomously build procedural memory. It can also spawn isolated subagents for parallel tasks and execute Python scripts via RPC to reduce context costs.
Who it’s for
- Power users who want a persistent AI assistant that lives across their CLI, Telegram, Discord, and other messaging apps.
- Developers looking for a research-ready agent capable of batch trajectory generation for training tool-calling models.
- Automation enthusiasts who need a natural-language cron scheduler for unattended tasks.
Highlights
- Closed Learning Loop: Automatically creates and improves skills from experience and nudges itself to persist knowledge.
- Multi-Platform Presence: Single gateway process providing continuity across Telegram, Discord, Slack, WhatsApp, Signal, and CLI.
- Flexible Deployment: Supports six terminal backends, including serverless options that hibernate when idle to save costs.
- Extensible Tooling: Integration with MCP (Model Context Protocol) servers and a built-in system of 40+ tools.
- Model Agnostic: Switch between providers and models with a simple command without changing code.