NousResearch/hermes-agent

The agent that grows with you

What it solves

Hermes Agent tackles the problem of AI assistants that lose context between sessions and never get better at their tasks. It is a self-improving agent that remembers what you tell it, learns skills from its own successful work, and keeps a growing model of who you are — so it gets more useful over time rather than starting from scratch.

How it works

Hermes runs a closed learning loop: after complex tasks it autonomously creates skills, those skills improve as they are reused, and it periodically nudges itself to save important knowledge. It searches its own past conversations (using FTS5 full-text search plus LLM summarization) to recall context across sessions, and uses the Honcho library to build a user profile. You can talk to it from a terminal UI, or connect it to messaging platforms like Telegram, Discord, Slack, WhatsApp, and Signal through a single gateway process. It supports any LLM provider (Nous Portal, OpenRouter, OpenAI, or your own endpoint) and can execute work on local machines, Docker, SSH, or serverless backends like Modal and Daytona, which hibernate when idle to save cost. It also has a built-in cron scheduler for automated tasks and can spawn isolated subagents for parallel work.

Who it’s for

Developers, researchers, and power users who want a persistent AI assistant that can run on a cloud server or VPS, be reached from messaging apps, and improve over time. It is also aimed at researchers generating training trajectories for tool-calling models.

Highlights

  • Built-in learning loop: creates skills from experience and improves them during use.
  • Persistent memory with cross-session search and LLM-based summarization.
  • Multi-platform access: CLI, Telegram, Discord, Slack, WhatsApp, Signal, and email via one gateway.
  • Runs on local, Docker, SSH, or serverless backends with near-zero idle cost.
  • Scheduled automations: natural-language cron jobs that deliver results to any platform.
  • Spawns parallel subagents and supports Python scripts calling tools via RPC.
  • Compatible with agentskills.io open standard and MCP servers.
  • Imports settings and memories from OpenClaw, Claude Code, Codex, and Cursor.

Related

  • Project
    fathah/hermes-desktop

    Hermes Desktop is an Electron‑based GUI that installs, configures, and lets you chat with the Hermes Agent LLM assistant. It supports local or remote back‑ends, dozens of LLM providers, a streaming chat UI with slash commands, profile & session management, tool integrations, scheduling, and many messaging‑gateway bots.

  • 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
    rlaope/oh-my-hermes

    An operating layer for Hermes Agent that adds structured workflow management, intelligent model routing, parallel task execution, and a persistent memory system to improve agent reliability and efficiency.

  • 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.

  • Project
    canwhite/Krebs

    A self-hosted AI gateway featuring a general-purpose agent with advanced context compression, goal-drift detection, and autonomous sub-agent management.