The short version

Pick OpenClaw if you want the biggest community and plugin ecosystem, or your life runs on iMessage and WhatsApp. Pick Hermes if persistent, long-term memory is the feature you care about, or you want one Nous Portal subscription to cover models and tools in a single bill.

You genuinely cannot go far wrong either way: both are open source, both talk to 20+ chat platforms, both run on cloud models so they need no GPU, and both install as an always-on background service. Both also move fast, so treat any comparison (including this one) as a snapshot: check their docs for this week's feature list.

What they are

OpenClaw started life as Clawdbot, became Moltbot, and settled on its current name in early 2026 while going properly viral. It's a personal AI assistant that lives on your machine, talks to you over your existing chat apps, and gets things done with real tool access. It runs on Node.js, ships iMessage, Telegram and a web chat in the core install, and adds Discord, Slack, WhatsApp, Signal and many more through an official plugin system. The community around it is enormous, and the canonical way to run it has always been a Mac mini that stays on.

Hermes Agent is Nous Research's take, released in February 2026 with the tagline "the agent that grows with you". Its headline feature is persistent memory: it accumulates context about you and your projects over time rather than starting fresh each session. It runs on Python, reaches 20+ messaging platforms through a single gateway, and pairs naturally with Nous Portal, a subscription that bundles 300+ models plus hosted tools (web search, image generation, text to speech, a cloud browser) behind one key. You can equally point it at OpenRouter, OpenAI, or your own endpoint.

🦞 OpenClaw at a glance

  • Runtime: Node.js 22+, installed with npm
  • Core channels: iMessage, Telegram, web chat
  • Plugin channels: WhatsApp, Signal, Slack, Discord, Teams, Matrix and more
  • Models: bring your own keys (Anthropic, OpenAI, others)
  • Always-on: launchd daemon via openclaw onboard --install-daemon
  • Stand-out: community momentum and the plugin ecosystem

☤ Hermes at a glance

  • Runtime: Python 3.11, installed with uv
  • Channels: 20+ through one gateway, including Telegram, Discord, Slack, WhatsApp, Signal, email and SMS
  • Models: Nous Portal (300+ models, one subscription), OpenRouter, OpenAI, or your own endpoint
  • Always-on: hermes gateway install
  • Stand-out: persistent memory that compounds over time

Side by side

OpenClawHermes
OriginCommunity project, previously Clawdbot and MoltbotNous Research, released February 2026
LicenceOpen source, free to runOpen source (MIT), free to run
RuntimeNode.js 22+ (npm)Python 3.11 (uv)
Model accessYour own provider keysNous Portal, OpenRouter, OpenAI, or your own endpoint
Chat channelsiMessage, Telegram and web chat in core; 20+ via official plugins20+ via a single gateway, configurable from a web dashboard
MemoryMarkdown workspace memoryPersistent memory as the headline feature
ExtensibilityPlugin and skill ecosystemHosted tool gateway via Nous Portal (search, images, speech, cloud browser)
GPU neededNone (cloud models)None (cloud models)
Always-on setupopenclaw onboard --install-daemonhermes gateway install

One practical note on iMessage: both agents reach it through a BlueBubbles server, which itself needs a Mac signed into iMessage. That's a genuinely macOS-only capability, and one of the reasons a Mac mini is the natural host for this whole category.

How to choose

  • Your life runs on WhatsApp or iMessage: OpenClaw treats iMessage as a core channel and WhatsApp as a first-party plugin, so it's the shortest path.
  • You want one bill for models and tools: Hermes with Nous Portal covers 300+ models plus search, images and speech under a single subscription.
  • You want maximum community and plugins: OpenClaw's ecosystem is the largest in the category right now.
  • You want an agent that remembers: long-term, compounding memory is what Hermes was built around.
  • You can't decide: they install side by side without conflict, or you can run one of each on two machines. Switching later is painless because your model keys are yours.

Either way, give it a machine of its own

An always-on agent doesn't belong on your laptop. Laptops sleep, reboot, change networks and leave the house. An agent that answers your messages at 3am needs a machine that's awake at 3am, on a connection that doesn't drop.

There's a security argument too. These agents are powerful precisely because they have real tool access: a shell, a filesystem, your connected accounts. That's a good reason to run one on dedicated hardware, away from your personal files, password manager and browser sessions. A single-tenant Mac mini in a rack, isolated from every other machine on its network, is a much better sandbox than the computer you do everything else on. That's exactly how we run ours: private by design, in UK racks, with nightly snapshots.

Both agents are available as bundles on any of our Macs. Pick OpenClaw assistant or Hermes AI agent when you order and the machine installs everything itself, usually within the hour, with a live progress tracker on your dashboard. When it hands over, you connect your own keys with openclaw onboard or hermes setup and your agent is home.

Give your agent a home.

A dedicated Apple silicon Mac with OpenClaw or Hermes preinstalled, online within the hour, from £39/mo. UK racks, private networking, cancel anytime.

Pick your Mac