It remembers
Persistent memory across sessions. Hermes keeps track of your preferences, your projects, and how it solved a problem last time — so you're not re-explaining yourself on every message.
Hermes runtime · Now available
Your Nous Hermes agent, hosted and online in minutes.
Hermes is NousResearch's open-source agent runtime, and it's been gaining ground in the open-agent community. Run it on Lobsterland with browser chat, Telegram, and Slack on an isolated instance — no servers, Docker, or Kubernetes to manage. Same account and billing as OpenClaw.
app.lobsterland.ing/app/chat
~2 minto a running agent
No public IPisolated by default
Your keysno model markup
One accountrun OpenClaw + Hermes
Not a stripped-down anything
Hermes is an agent that grows with you.
NousResearch built Hermes around a learning loop most agents don't have: it remembers, it writes its own skills, and it gets sharper the longer it runs. That's why people choose Hermes on its own terms — not as a lighter OpenClaw.
It improves itself
After a hard task, Hermes writes its own reusable skill and sharpens it with use. Nous designed it to get more capable the longer it runs.
It's open, and it's yours
An open-source agent from NousResearch, model-agnostic by design: bring your own keys for Anthropic, OpenAI, Google, or OpenRouter — which also serves Nous's own open-weight Hermes models. No lock-in.
A persistent agent, not a chatbot
Hermes is autonomous and long-lived: it lives on your isolated instance, works across your channels, and carries its memory between conversations.
Why host Hermes here
A real Hermes agent, without the ops.
You bring the agent you want to run; we handle the rest — provisioning, isolation, channels, and logs. No infrastructure to stand up, no CLI to babysit.
Chat the moment it launches
Message your Hermes agent right from the Lobsterland dashboard — nothing to install, nothing to wire up.
Telegram & Slack, with real access control
Allowlist exactly who can message it — specific people, a whole channel, or anyone — and require an @mention in groups so it only replies when called.
Bring your own model keys
Run on Anthropic, OpenAI, OpenRouter, or Google. No markup — you pay your provider directly, and keys stay encrypted on your instance.
Isolated & zero-ops
Each agent is its own isolated instance with no public IP. Watch what it does in the logs; we keep the runtime patched and online.
From zero to running
Three steps to a live Hermes agent.
Step 1
Sign in and pick Hermes
Choose the Hermes runtime when you launch an instance — same signup as OpenClaw.
Step 2
Add a model key and your channels
Drop in an Anthropic, OpenAI, OpenRouter, or Google key, then connect Telegram and Slack with allowlists.
Step 3
Start chatting
Talk to it in the browser or from your channels. It's online in about two minutes.
Two different agents, one account
Hermes or OpenClaw — pick the agent, not a tier.
They aren't the same product with features taken away. Hermes is the self-improving chat-and-channels agent; OpenClaw is the full control plane. Run either or both on one account, and switch experiments anytime.
Reach for Hermes
A self-improving agent that lives in chat — browser, Telegram, and Slack. It remembers across sessions, writes its own skills, and runs on whatever model you bring.
Reach for OpenClaw
The full operating surface: workspace files, skills you manage, cron, multiple agents, Hosted Browser, Paperclip, usage analytics, and WhatsApp.
FAQ
Hermes hosting, answered.
Hermes is NousResearch's open-source agent runtime — "the agent that grows with you." On Lobsterland you get a managed Hermes instance with browser chat, Telegram, Slack, and runtime logs, running on an isolated instance with no servers, Docker, or Kubernetes to maintain — on the same account, dashboard, and billing as OpenClaw.
Hermes is a different kind of agent, not a lighter OpenClaw. It's built around a learning loop: it keeps persistent memory across sessions, writes its own skills after solving hard tasks, and gets more capable the longer it runs. It's open-source from NousResearch and model-agnostic — bring your own keys, and through OpenRouter you can even run Nous's own open-weight Hermes models. OpenClaw is the broader operating surface (workspace, cron, multi-agent routing, Hosted Browser, Paperclip); Hermes is the self-improving chat-and-channels agent.
Bring your own model keys from Anthropic, OpenAI, OpenRouter, or Google — and through OpenRouter you can run NousResearch's own open-weight Hermes models. There's no markup: you pay your model provider directly, and your keys stay encrypted on your isolated instance.
You decide. Allowlist specific Telegram and Slack users, open it to a whole channel, or allow anyone — and require an @mention in group chats so the agent only replies when it's called.
Yes. Pick the runtime when you launch an instance and run either — or both — on one account and one bill. It's the simple way to evaluate runtimes without re-onboarding or a second set of keys.
Hermes runs on the same capacity tiers as OpenClaw: Light is $6.90/month, Medium is $14.90/month, and Large is $29.90/month. Model usage is billed by your own provider.
Launch your Hermes agent today.
Your Nous Hermes agent — running in about two minutes, on your own model keys, with allowlisted Telegram and Slack. No servers, no markup, from $6.90/mo.