NousResearch/hermes-agent

The agent that grows with you

What it solves

Hermes Agent is a self-improving AI agent designed to overcome the limitations of static AI assistants by implementing a closed learning loop. It solves the problem of "forgetting" across sessions and the need for manual tool configuration by autonomously creating and refining skills from experience and building a deepening model of the user over time.

How it works

The agent operates through a learning loop where it curates its own memory, creates new skills after completing complex tasks, and improves those skills during use. It uses FTS5 session search with LLM summarization for cross-session recall and integrates with the Honcho dialectic user modeling system.

Users can interact with it via a terminal interface (TUI) or a messaging gateway that connects to platforms like Telegram, Discord, Slack, WhatsApp, and Signal. It is model-agnostic, allowing users to switch between providers (e.g., Nous Portal, OpenRouter, OpenAI) without code changes. For execution, it supports multiple terminal backends including local, Docker, SSH, and serverless options like Modal and Daytona.

Who it’s for

  • Power users and developers who want a persistent, cross-platform AI assistant that can be hosted on a VPS or GPU cluster.
  • Researchers looking to generate and compress trajectories for training tool-calling models.
  • Users seeking automation who need scheduled, natural-language tasks (cron jobs) delivered to messaging apps.

Highlights

  • Closed Learning Loop: Autonomous skill creation and self-improvement based on experience.
  • Multi-Platform Gateway: Single process to manage conversations across Telegram, Discord, Slack, and more.
  • Flexible Deployment: Runs on everything from a $5 VPS to serverless infrastructure that hibernates when idle.
  • Model Agnostic: Easy switching between LLM providers via simple commands.
  • Advanced Tooling: Built-in cron scheduler, parallel subagent spawning, and MCP (Model Context Protocol) integration.

Related

  • Project
    outsourc-e/hermes-workspace

    A comprehensive command center for AI agents that integrates chat, memory, skills, and a terminal into a single workspace for orchestrating multi-agent swarms.

  • Project
    EverMind-AI/Raven

    Raven is an open-source, self-improving agent harness that integrates long-term memory, proactivity, and local tracing for long-running AI agent workflows.

  • Project
    fuyuxiang/echo-agent

    A self-hosted, long-running AI agent featuring a four-layer cognitive memory system and a self-evolution engine to improve skills over time.

  • Project
    MemTensor/MemOS

    MemOS 2.0 is a Memory Operating System for LLM‑based AI agents. It offers a unified API to store, retrieve, edit, and delete multi‑modal memories (text, images, tool traces) organized as a graph. The system can run as a hosted cloud service (API‑key access) or be self‑hosted via Docker (Neo4j + Qdrant) or a local SQLite plugin. Ready‑made plugins integrate MemOS with OpenClaw, Hermes, and DeepSeek Harness. Benchmarks show it outperforms existing commercial memory products. Licensed under Apache 2.0.

  • Project
    Memento-Teams/Memento-Skills

    A skill-centric agent runtime that allows AI agents to design, execute, and persist reusable skills to improve their performance on complex tasks through a reflection-based learning loop.