This is an early release preview. You may encounter bugs.
Hermes logo
Unclaimed

Harness memory connectors

Hermes

Self-improving AI agent with a learning loop that creates and refines skills, recalls past sessions, and runs across chat platforms

A 86/100 GitHub score ? This grade is derived from GitHub signals, not user votes. Open for the full breakdown.
100% recommend · 1 rater

01 / About

What Hermes is.

Hermes Agent is an open-source agent harness from Nous Research built around a learning loop: it creates skills from experience after complex tasks, refines them during use, nudges itself to persist knowledge to memory, and searches its own past conversations so what it learns in one session carries into the next. You run it, choose a model, and talk to it from a terminal or a chat app.

It works with any model provider (Nous Portal, OpenRouter, OpenAI, a self-hosted endpoint, and others) and switches with hermes model or /model without code changes. One gateway process serves Telegram, Discord, Slack, WhatsApp, Signal, email, and Home Assistant, with conversations continuing across platforms and voice memos transcribed on arrival.

Tools run on one of seven terminal backends: local, Docker, SSH, Singularity, Modal, Daytona, and Vercel Sandbox. Daytona and Modal add serverless persistence, so the agent's environment hibernates when idle and wakes on demand. A built-in cron scheduler runs natural-language automations unattended and delivers results to any connected platform.

Beyond interactive use, Hermes can spawn isolated subagents for parallel workstreams, let the model write Python scripts that call tools over RPC to collapse multi-step pipelines into a single turn, and generate and compress batches of trajectories for training tool-calling models. Nous Portal optionally bundles 300+ models and a tool gateway (web search, image generation, text-to-speech, cloud browser) behind one login.

Features

  • Learning loop: autonomous skill creation and self-improvement, agent-curated memory with periodic nudges, and FTS5 session search with LLM summaries for cross-session recall
  • Terminal UI: multiline editing, slash-command autocomplete, conversation history, interrupt-and-redirect, and streaming tool output
  • Messaging gateway: Telegram, Discord, Slack, WhatsApp, Signal, email, and Home Assistant from one process
  • Provider-agnostic models: switch providers and models per conversation with /model provider:model
  • Scheduled automations: cron jobs written in natural language with delivery to any platform
  • Subagents and RPC scripting: parallel isolated subagents and tool calls from Python scripts
  • Execution backends: local, Docker, SSH, Singularity, Modal, Daytona, and Vercel Sandbox
  • Tools and toolsets: 40+ built-in tools, a toolset system, MCP server integration, and project context files
  • Security controls: command approval, DM pairing, and container isolation
  • Personalities and sessions: /personality, /retry, /undo, /compress, /usage, and /insights shared across CLI and chat
  • Platform coverage: Linux, macOS, WSL2, Termux on Android, and native Windows, plus a desktop app
  • OpenClaw migration: hermes claw migrate imports an existing OpenClaw setup

02 / Discussion CREDIBILITY-GATED

Discussion

Reading is open to everyone. Posting and voting need a verified identity or a GitHub grade of B or higher.

  • No discussions yet.

04 / Build

Build with Hermes.

Browse the catalogue for frameworks, tools, and harnesses, each scored on real GitHub credibility.

Get Hermes →

Browse the catalogue