← Back to all projects

hermes-agent

A self-hosted personal agent that writes its own memory and skills, reachable from a TUI or Telegram.

OfficialMIT
Stars
259k
Forks
56.9k
Open issues
52.1k
Last commit
30 Oct 2026

What is hermes-agent?

Hermes Agent is Nous Research's own agent: a terminal UI plus a single gateway process that carries the same conversation into Telegram, Discord, Slack, WhatsApp, Signal and Email, with a cron scheduler, isolated subagents, and seven terminal backends from local and Docker to Modal and Vercel Sandbox. Its distinguishing piece is a closed learning loop — the agent curates its own memory on periodic nudges, creates skills after complex tasks, and searches past sessions through FTS5 with LLM summarization — which means the state that makes it useful accumulates under `~/.hermes`, outside your version control, and needs occasional pruning. It is an assistant you run and configure rather than a library you build on: the README documents `hermes` subcommands and slash commands, not an embedding API, so it is a poor fit if you wanted an agent loop to call from your own code. Putting a shell-capable agent behind chat platforms also makes the command-approval and DM-pairing settings load-bearing rather than optional.

What can you do with hermes-agent?

  • Run it on a server and talk to it from your phone — hermes gateway setup then hermes gateway start serves Telegram, Discord, Slack, WhatsApp, Signal and Email from one process, with voice memo transcription and conversation continuity across platforms. /model, /compress and /undo work on both sides, but the two interfaces are not identical: usage insights are /insights [--days N] in the CLI versus /insights [days] in chat, and interrupting is Ctrl+C in the TUI versus /stop in a message.
  • Let the environment hibernate between sessions — Terminal backends cover local, Docker, SSH, Singularity, Modal, Daytona and Vercel Sandbox. Daytona and Modal add serverless persistence: the agent's environment hibernates when idle and wakes on demand, so an idle deployment costs close to nothing.
  • Change model or tool provider without touching code — hermes model selects provider and model — OpenRouter, OpenAI, Nous Portal or your own endpoint. hermes setup --portal runs the OAuth login and enables the Tool Gateway, routing web search (Firecrawl), image generation (FAL), TTS (OpenAI) and a cloud browser (Browser Use) through one subscription; the gateway is per-backend, so you can still supply your own key per tool. hermes portal info prints what is wired up.
  • Carry an OpenClaw setup across — hermes claw migrate imports SOUL.md, MEMORY.md and USER.md entries, user-created skills into ~/.hermes/skills/openclaw-imports/, the command allowlist, messaging configs and allowlisted API keys. --dry-run previews the plan and --preset user-data migrates without secrets; hermes setup also detects ~/.openclaw on first run.
  • Schedule unattended runs and fan work out to subagents — The built-in cron scheduler takes tasks written in natural language and delivers results to any connected platform; the setup is documented under docs/user-guide/features/cron. Parallel workstreams go to isolated subagents, and Python scripts that call tools over RPC collapse a multi-step pipeline into one zero-context-cost turn. Note that the README's command list exposes no scheduling subcommand — the entry point is the conversation itself.

Before you choose hermes-agent

  • Changing provider does not change the shape of the bill. An analysis puts roughly 13,900 tokens on every call before any of your conversation is processed, most of it tool definitions and the system prompt.Reported in#4379

Star history

16 Aug to 30 Oct · +27.4k

231k259k

Frequently asked questions

Is hermes-agent free for commercial use?

hermes-agent is released under the MIT licence — OSI-approved open source, which permits commercial use.

How can hermes-agent be deployed?

hermes-agent is available as Self-hosted / Runs locally.