Run Two AI Agents on One Instance: OpenClaw + Hermes-AI Co-Hosting

OpenClaw Box now runs two agent runtimes. Alongside OpenClaw — the open-source assistant platform you already know — you can now deploy Hermes-AI, either on its own or co-hosted right next to your OpenClaw assistant on the same instance. Same 60-second Telegram deploy, same managed models, same cloud Chrome.

This post covers what changed, why co-hosting matters, and the one command that turns a single-agent instance into a two-agent team.

Two runtimes, one managed instance

Until now, an OpenClaw Box instance ran a single runtime: OpenClaw. That is still the default and still what most people want. But a single assistant is not always the right shape. Sometimes you want a second agent with a different model, a different personality, or a different job — without paying for and managing a second instance.

So we added a second supported runtime: Hermes-AI (built on NousResearch's hermes-agent). You can now:

• Deploy a standalone Hermes-AI instance — pick Hermes at deploy time instead of OpenClaw.
• Co-host Hermes alongside OpenClaw — run both agents on the same VM, sharing resources.

Both options are fully managed. No servers, no Docker, no glue code.

Why co-hosting

The interesting part is co-hosting. When you add Hermes to an existing OpenClaw instance, the two agents share the expensive infrastructure:

• One cloud Chrome. Both agents attach to the same browser over CDP — your logins and sessions are available to both, and you only pay for one browser.
• One model gateway. Both runtimes route through the same LiteLLM proxy, so your hosted models and credits are shared.
• One claude-max subscription. If you've wired up claude-max, both agents reuse it.

What stays separate is what should: each agent has its own Telegram bot, so you talk to them independently. Your OpenClaw assistant keeps doing its thing on @OpenClawBoxBot while Hermes answers on its own bot — no crossed wires.

The one command

If you already have a VM instance, adding Hermes is a single Telegram command:

/add hermes <token>

Create a bot with @BotFather, grab its token, and send it to your OpenClaw Box bot with /add hermes. Behind the scenes we install the Hermes runtime onto your existing VM, point it at your shared browser and model gateway, and bring up a second gateway on its own port. When it's ready, you get back the new bot's username — message it directly and you're talking to Hermes.

Prefer the web console? The same flow is available there: open your instance and use the Add a Hermes agent card.

Who this is for

• You want a second opinion. Run OpenClaw and Hermes side by side on the same data and compare how they reason.
• You want specialized agents. Keep OpenClaw as your generalist assistant and give Hermes a focused role.
• You're cost-conscious. Co-hosting means a second agent without a second instance bill — they share the browser, models, and VM.

Get started

If you're new, deploy OpenClaw in 60 seconds: open @OpenClawBoxBot and type /create. Want Hermes-AI as your primary runtime? Pick it at deploy time in the web console. And if you already run an instance, add a co-hosted Hermes agent today with /add hermes <token>.

Two runtimes. One instance. Your choice.