Hermes Agentvs Muse
Hermes Agent runs on your own hardware or Hermes Cloud and works with the model you pick, Meta's Muse Spark included. Muse runs in Meta's cloud on Meta's own models.
- Your machines or Hermes Cloud
- Meta's cloud
- Hundreds, Muse Spark included
- Meta's Muse models
- More than 20 platforms
- Anywhere you install it
- Five countries, limited testing
Why Hermes Agent
Runs on your hardware or Hermes Cloud
Install it on a computer, a server or a VPS, or use Hermes Cloud: every paid Nous Portal plan includes one Hermes Cloud agent. Muse runs on Meta's cloud.
Hundreds of models, Muse Spark included
Nous Portal adds frontier models on launch day, and OpenRouter, local models and Muse Spark via the Meta Model API work too. Muse runs on Meta's own models.
Answers on more than 20 messaging platforms
Telegram, Discord, Slack, WhatsApp, Signal, iMessage and Microsoft Teams connect through one gateway. Muse works in its own apps, on the web and in WhatsApp.
Open source under the MIT license
You can inspect and change the code, and the license allows commercial use. Meta keeps all rights in Muse, and its terms forbid reverse engineering it.
At a glance
| Topic | Hermes Agent | Muse |
|---|---|---|
| Source code | Edge: Open source under the MIT license, which allows commercial use. | Closed source. The terms forbid reverse engineering. |
| Where it runs | Edge: Your computer, a server, a VPS or Hermes Cloud. | A persistent Linux computer in Meta's cloud for each Muse. |
| Models | Edge: Hundreds via Nous Portal, plus OpenRouter, local models and Muse Spark. | Meta's own models: Muse Spark, and Muse Wonder on paid plans. |
| Chat apps | Edge: More than 20 messaging platforms, including Telegram, Slack and WhatsApp. | The Muse apps, the web, and WhatsApp once you link your account. |
| Your data | Edge: Stored on the machine that runs it. No telemetry by default. | Kept on a Meta-hosted cloud computer. |
| Where you can use it | Edge: Anywhere you install it, or Hermes Cloud in the nearest available region. | US, Canada, Japan, Singapore and Mexico, for adults 18+, in limited testing. |
| Multiple agents | Edge: Up to 10 subagents at once by default, and Bot Mode chats of 2 to 6 Bots. | One personal agent with side chats. No agent teams you set up. |
Switch from Muse
Install Hermes Agent
Get Hermes Desktop for macOS (Apple Silicon), Windows and Linux, or the terminal installer on a server or a VPS.
Run hermes setup --portal
One command signs you in to Nous Portal, picks a model and turns on the Tool Gateway. Add a Meta Model API key for Muse Spark.
Keep a copy of your Muse data
In Muse, open Settings, then Data controls, and download a copy of your chats, files and other agent information.
hermes setup --portalIn detail
It runs on your hardware or Hermes Cloud
Muse is Meta's personal AI agent. Meta gives each Muse its own cloud computer, the Muse Secure VM: a persistent, isolated Linux computer with a browser that you and the agent share. Muse is for adults 18 and over with a Meta account, and Meta says Muse and its subscriptions are still in limited testing and not available in all locations.
Hermes Agent runs wherever you install it: your computer, a server or a VPS. Hermes Desktop runs on macOS (Apple Silicon), Windows and Linux. Its shell commands can run locally, in a Docker container, over SSH, or in Modal, Daytona or Vercel Sandbox.
If you do not want to keep a machine on, Hermes Cloud hosts your agent on a dedicated, always-on instance in the nearest available region. Every paid Nous Portal plan includes one Hermes Cloud agent; larger or extra agents bill daily from credits. Hermes Desktop can connect to that agent and to agents on your own machines, side by side.
You pick the model, Muse Spark included
Hermes Agent works with any OpenAI-compatible API. One Nous Portal login covers hundreds of models and hosted tools, and the provider list includes OpenRouter and dozens of others. A ChatGPT sign-in goes through the Codex provider and uses Codex models. Local models run through Ollama, vLLM, llama.cpp or SGLang with no account.
Meta's Muse Spark models work too, through the Meta Model API provider and your own Meta Model API key. The model you pick is the model that answers: /model switches mid-session, your conversation history, memory and skills carry over, and backup providers are opt-in.
Muse runs on Meta's own models. On October 21, 2026, Meta's CEO said Muse Spark is the default on the free tier and that paid subscribers can use Meta's newer Muse Wonder model. Meta documents no way to pick another company's model or use your own API key in Muse, and it counts usage in Muse tokens at rates it sets.
It stores your data on the machine you choose
Hermes Agent stores conversations, memory and skills in the Hermes home directory on the machine that runs it (~/.hermes on macOS and Linux). Model requests go only to the providers you configure. Hermes Agent collects no telemetry by default; usage-metrics sharing is opt-in, and those metrics leave out prompts and responses.
An encrypted vault on your machine fills saved logins and cards without the model seeing them, and it asks you before every card fill. The vault works with 1Password and Bitwarden.
Muse keeps your data on its Meta-hosted cloud computer. Meta says Muse keeps your conversations and that computer's data out of its ad systems. Meta limits staff access by policy but says its current design does not prevent it from accessing data when necessary to support, secure or operate the service. Meta plans a Confidential VM for later in 2026.
It answers on more than 20 messaging platforms
Through one gateway, Hermes Agent connects to more than 20 messaging platforms, including Telegram, Discord, Slack, WhatsApp, Signal, iMessage through BlueBubbles, Microsoft Teams, LINE and email. Scheduled jobs can post their results to those chats. The CLI, Desktop, bots and scheduled jobs share one gateway, so a conversation continues wherever you pick it up.
A Hermes Agent mobile app is in beta and coming soon. Today you message it from your phone through a chat app, or use Hermes Desktop in your phone's browser with the Webapp.
Meta lists the Muse app, WhatsApp, the web and the Mac app as the ways to reach Muse. WhatsApp works once you link your WhatsApp account, and Muse sees only its own chat there.
It runs teams of agents you set up
Hermes Agent can hand parts of a task to subagents, each with its own context and terminal session, and it runs up to 10 at once by default. Profiles let you run several independent agents on one machine, each with its own memory, skills and API keys.
In Hermes Desktop, Bot Mode gives each Bot its own role, model, memory and skills. Bots run recurring routines, work together in group chats of 2 to 6, and message each other directly.
Muse is a personal agent with a main chat and side chats. It starts subagents internally, and Meta documents no agent teams you set up.
Its code is open source under the MIT license
Hermes Agent is open source under the MIT license, which allows commercial use, and the code is on GitHub. Its skills are compatible with the agentskills.io open standard, and you can share the skills it writes or install community skills from the Skills Hub.
Muse is closed source. Meta keeps all rights in it, and the Muse terms forbid reverse engineering it. In October 2026 Meta released open weights for Muse Spark Open, a version of its Muse Spark model, under the Muse Model License, which has limited commercial terms.
Sources (24), checked November 9, 2026
Every statement about Muse on this page comes from the pages below. If something is out of date, tell us at support@nousresearch.com and we will correct it. Statements about Hermes Agent come from its documentation, source code and Nous Portal.
- Introducing Muse: The World's First Personal AI Agent Built for Everyone (Meta Newsroom)
- Muse Supplemental Terms of Service
- The Future Is for Everyone: Muse for Small Business (Meta Newsroom)
- Introducing Muse Spark 1.3 (Meta AI Research)
- Muse Spark Is Now an Open-Weights Model (PyTorch blog)
- Introducing Muse Spark Open, The Most Capable US Open-Weight Model to Date (Meta AI Research)
- How We Built Safety Into Muse (Meta AI Research)
- Muse: Meta's personal AI agent, features and capabilities
- Everything We Announced at Meta Connect 2026 (Meta)
- How Muse works with files and apps in your Mac (Meta Help Center)
- Introducing Muse Wonder and Muse Wonder Full: Frontier Models for Agents (Meta Newsroom)
- How to buy additional Muse tokens (Meta Help Center)
- How to manage your Muse data (Meta Help Center)
- Muse Privacy Policy
- How Muse works with skills (Meta Help Center)
- Muse for Small Business: Build Your Company with AI
- How to get started with Muse (Meta Help Center)
- How to connect or disconnect your WhatsApp account from Muse (WhatsApp Help Center)
- Subscribe to Muse (Meta Help Center)
- Muse Is Now Available in Singapore (Meta Newsroom)
- Presentamos Muse, with the November 5 update on Mexico (Meta Latin America newsroom, in Spanish)
- Meta launches Muse in Japan (Meta Japan newsroom, in Japanese)
- About Muse subscriptions (Meta Help Center)
- Meta Q3 2026 Earnings Call Transcript (Meta Investor Relations)
