We Hosted OpenClaw So You Don't Have To

#ai #openclaw #api #devops

TLDR: we just launched free OpenClaw hosting. One-click deploy, no infrastructure to manage, only pay for the AI tokens your agent uses.

Free Hosted OpenClaw for Everyone
What Is OpenClaw?
How Provisioning Works
What You're Actually Running
Model Switching Without Redeployment
Pricing Model
Current Limitations
FAQ

Free Hosted OpenClaw for Everyone

Self-hosting OpenClaw is certainly doable (if anything, a bit of a faff).

The community guide walks you through provisioning a VPS or setting up Docker, creating SSH keys, connecting to the server, installing Node.js and dependencies, installing OpenClaw itself, configuring it, setting up TLS certificates, adding a TLS renewal cron job, connecting an LLM provider, and pairing a messaging channel. The guide even has time estimates for each step. The total is about 80 minutes before your first message goes through, and that assumes you're familiar with all the aforementioned tech, and nothing breaks along the way. After that, ongoing updates, monitoring, and troubleshooting are all yours.

That might be reasonable if you need full control over your infrastructure. If you don't, it's a lot of overhead for something that should just work.

We've built a free hosted version of OpenClaw on Runware to take the stress out of the whole process. This post covers how the setup works, what the infrastructure looks like, and where the current limits are.

What Is OpenClaw?

OpenClaw 🦞 is an open-source framework for running AI agents. The key word is agent: it doesn't just answer questions, it actually does things. Clear your inbox, send emails, manage your calendar, check you in for flights, browse the web, run shell commands, control your browser, write and execute code, the possibilities are endless.

You connect it to channels such as WhatsApp, Telegram, Discord, Slack, Signal, iMessage, or similar, and from that point on you interact with it the same way you'd message anyone else. It has persistent memory, supports community-built skills, and can write new skills for itself.

It's incredibly useful and has exploded in popularity since launch. We believe everyone should be able to get in on the action, without the stress of a self-hosted deployment.

How Provisioning Works

On Runware, when you click "Deploy OpenClaw" from the OpenClaw dashboard, the platform spins up a container for OpenClaw, in an isolated secure environment. This takes about three minutes.

The part that usually causes friction in a manual setup is connecting your LLM provider. You'd normally generate an API key from your provider, copy it, SSH into your server, and paste it into a config file. With the hosted version, that step is removed entirely. Runware generates an API key tied to your Runware account and pulls it directly into your OpenClaw instance. By the time the container is live, inference is already wired up. You don't see a key, you don't have to paste anything, and there's no config file to edit.

From there you open OpenClaw's Control UI, pick a model, connect a messaging channel, and you're running.

What We're Actually Running

The hosted container we're deploying is the same OpenClaw you'd run yourself, deployed into a isolated environment on our infrastructure. Updates to the container and underlying infrastructure are handled automatically.

Channels are configured through the Control UI after deployment. Supported channels include WhatsApp, Telegram, Discord, Slack, Signal, Google Chat, iMessage, and Nostr.

Model Switching Without Redeployment

Because inference routes through our hardware, you can switch between models without redeploying anything - it's as simple as selecting a new model.

We're offering the full Runware LLM catalogue, including both open-source and closed-source models, including

GLM-5.1 & GLM-4.7
Qwen3.5-397B & Qwen3.5-27B
Kimi K2.6
MiniMax M2.7, MiniMax M2.7 Highspeed & MiniMax M2.5
DeepSeek V4 Flash
Claude Opus 4.7, Claude Sonnet 4.6 & Claude Haiku 4.5
GPT-5.5, GPT-5.4, GPT-5.4 Mini & GPT-5.4 Nano
Gemini 3.1 Pro, Gemini 3.1 Flash Lite, & Gemini 3 Flash.
Grok 4.3

New models added to the Runware model catalogue will show up for use, automatically.

Pricing

Hosting OpenClaw on Runware is entirely free. Compute, storage, networking, the container, updates, and security are all included at no charge. There's no subscription, no monthly minimum spend, and no per-instance fee. All you need is credit in your Runware account to cover LLM usage.

Model inference is billed at our market leading rates; you're only charged for LLM token usage - a fully pay-as-you-go model. New accounts with a business email get $2 in free credit to start with.

Current Limitations

This system is in beta, and there's a few things to know before you start building;

Each Runware account is currently limited to one instance. If you need multiple agents running in parallel, the self-hosted path is the better option for now.

We're monitoring performance and listening to community feedback as the service matures. If you run into issues, or want to share what you've built, we'd love to hear from you in the Runware Discord.

Get started

Happy to answer questions in the comments about anything that's unclear, and we'd love your suggestions and input!

You can also join the conversation and stay up to date with the latest news in the Runware Community Discord

FAQ

How fast can I deploy OpenClaw?
The provisioning itself takes about three minutes. From clicking deploy to having a working agent is under five minutes total, and most of that is the sign-up flow. There's no configuration to fill in and nothing to install locally.

Is hosting really free?
Yes. Compute, storage, the container, networking, updates, and security are all included at zero charge. The way Runware's model works is that you pay for the AI inference your agent generates, not for the infrastructure running it. If your agent is live but never calls a model, you pay nothing.

How does pricing work?
The only thing you pay for is the AI tokens your agent uses. Open-source models are billed at Runware's published rates. There's no monthly subscription and no per-instance charge on top of that.

What models does OpenClaw support on Runware?
Every model in the Runware catalogue is available, and you switch between them from inside OpenClaw without redeploying. That covers open-source models like Qwen, GLM, Kimi, MiniMax, and DeepSeek running on Runware's own infrastructure, and closed-source frontier models including Claude, GPT-5, Gemini, and Grok. New models are added to your instance automatically as they're added to the catalogue.

How do updates work?
We handle updates to the OpenClaw container and the underlying infrastructure automatically.

Is my instance isolated?
Yes. Each OpenClaw instance runs in its own secure environment inside an isolated container. There's no shared state with other customers' instances.

What happens if I outgrow the hosted tier?
OpenClaw is open source, so you can move to a self-hosted setup at any point. Nothing about the hosted setup locks you into any ongoing contract.

What happens if I don't use my instance for a while?
If an instance is unused for 5 days it's temporarily suspended to free up capacity. You won't lose any data. Just log back into the Runware dashboard and redeploy, it takes a couple of minutes to reinitialize, then you continue as normal.