OpenClaw connects to cloud-hosted AI models by sending each request to the model’s provider with an API key that identifies your account and authorizes the call. With Bluehost AI Gateway, the connection works like this:

• One key for multiple providers: You get a single API key that works across the providers we support, so you don’t need a separate account with each one.
• One wallet for usage: Every request draws from one prepaid AI Credits wallet, and all usage appears on one bill.
• Quick setup: You generate the key in your control panel and add it to OpenClaw’s settings on your Self-Managed VPS.
• No key needed for local models: Models running through Ollama on your server use your own hardware, so they don’t need a key or credits.

OpenClaw is an open-source, self-hosted AI agent, and our guide to what OpenClaw AI is covers the basics if you are new to it. Below, we explain how the connection works, what it costs and how to set it up securely.

OpenClaw’s gateway vs. Bluehost AI Gateway: What’s the difference?

OpenClaw’s internal gateway and Bluehost AI Gateway share a name, but they sit at different layers of your setup and solve different problems.

• OpenClaw’s internal gateway: This is OpenClaw’s own control plane. It decides where each request goes by routing it to a model.
• Bluehost AI Gateway: This product handles how you authenticate and pay for model access, through one API key and prepaid AI Credits.

The two work together. OpenClaw’s gateway picks the destination, and our AI Gateway provides the key that lets the request through.

Why is connecting to models usually the hard part?

Normally, connecting an agent to models means signing up with each provider separately. For every provider you want to use, you have to:

• Create an account
• Generate an API key
• Paste the key into your app
• Track a separate bill

That work tends to land at the exact moment you are trying to get the app running.

Costs are also hard to predict, because LLM providers charge by usage. The OpenAI Help Center explains how LLM usage is measured in tokens, with input and output tokens priced separately.

Spread that pricing across several invoices and it gets difficult to see what you are spending. Consolidating usage into one bill makes that spending easier to read.

Our AI Gateway removes both problems by replacing multiple provider accounts and invoices with one key, one wallet and one bill.

How Bluehost AI Gateway simplifies model access?

Our AI Gateway replaces separate provider accounts with one wallet, one bill and API keys that work across all the providers we support. Here’s what that means in practice:

• No provider sign-ups: You reach every supported provider through our AI Gateway, so you don’t need an account with each one.
• One key across models: A single API key works across all supported models, so you don’t juggle a different key per provider.
• One prepaid wallet: You buy AI Credits up front, and all usage draws from the same balance.
• One bill: All usage appears on a single bill from us instead of separate provider invoices.
• Works across your servers: The wallet sits at the account level, so it’s shared across all your Self-Managed VPS and VDS instances rather than tied to one server.

Tip: You can create as many keys as you need, and each one draws from the same wallet. We suggest one descriptively named key per app so usage stays easy to track.

How AI Credits and usage work?

AI Credits are prepaid and priced at a flat rate across all packs. You buy a balance up front, and there is no subscription or auto-renew. The packs available today are:

Pack	Credits	Price	Rate per credit
Starter	500	$5	$0.010
Basic	1,000	$10	$0.010
Enhanced	5,000	$50	$0.010
Ultra	10,000	$100	$0.010

Pack	Credits	Price	Rate per credit
Starter	500	$5	$0.010
Basic	1,000	$10	$0.010
Enhanced	5,000	$50	$0.010
Ultra	10,000	$100	$0.010

Note: Pricing is as of September 2026. For the latest pricing and offerings, visit Bluehost AI Gateway | Simplify LLM Access With One API Key.

Credits are non-refundable and stay valid for 12 months from your most recent purchase. Buying more tops up the same wallet and resets the validity date for the whole balance. Because all keys draw from that shared balance, your usage is limited by the credits you buy.

With the wallet mechanics covered, adding to that balance takes one click.

What’s included in Bluehost AI Gateway, and what isn’t yet?

Bluehost AI Gateway brings provider access, AI Credits and billing together into one relationship with us. It’s available on Self-Managed VPS and Virtual Dedicated Server (VDS) plans. Some features are still on the way, so here’s what you can expect in this first release.

What’s included

You get access to all the providers we support through one prepaid wallet, API keys you create yourself and a single bill. The wallet is shared across all your Self-Managed VPS and VDS instances, so you manage credits in one place no matter how many servers you run.

Current limitations

A few features aren’t available yet, so it helps to know them before you set up:

• Manual setup: You create an API key in your control panel and paste it into OpenClaw yourself. It’s a one-time step per app.
• No bring-your-own-key (BYOK): You can’t use your existing OpenAI or Anthropic keys through the gateway. All usage runs on AI Credits.
• No advanced usage controls: Per-model routing rules, per-agent spending caps and multi-team allocations aren’t supported yet.
• VPS and VDS only: The gateway isn’t available on shared hosting or managed WordPress plans.

With these limits in mind, the setup itself is straightforward and takes only a few steps.

How to connect OpenClaw to AI models with Bluehost AI Gateway?

Here is the end-to-end path once you have a self-managed VPS running OpenClaw. If you still need to deploy the app itself, our guide to set up OpenClaw on a VPS walks through the full one-click install.

1. Add AI Credits: Buy a pack at the VPS/Virtual Dedicated Server Hosting (VDS) Configurator when you select an eligible AI app.
2. Create your API key: Open the API Keys area and create a key. It is shown once, so copy it now and name it descriptively.
3. Add the key to OpenClaw: Paste it into OpenClaw’s model-provider settings. Our OpenClaw configuration guide shows the exact fields to use.
4. Choose a model and test: Point OpenClaw at a supported model by naming the provider and model, then run a test request.

Existing customers can skip the Configurator and add credits from the AI Tooling tab on a self-managed VPS or Virtual Dedicated Server Hosting (VDS) plan. That works for both a first-time purchase and a top-up.

There is no forced default model, so start with a general-purpose model from a provider on our supported list. You name that provider and model in OpenClaw’s model-provider settings.

OpenClaw can connect to cloud providers such as OpenAI or Anthropic Claude, or to local models via Ollama.

Keeping your API keys and data secure

Your API key is a secret, so treat it like a password. Under the IETF bearer token standard, any party holding the token can use it. Because every key draws from your shared, account-level wallet, a leaked key can spend credits meant for all your apps and servers.

A few habits help you avoid that from the start:

• Store keys in a secrets manager: Keep them in a vault or environment variable rather than in code or config files you commit to a repository.
• Use a separate key for each app: If one key is exposed, you can revoke it without disrupting your other apps.

For broader guidance on securing OpenClaw itself, including credential handling and tool permissions, see our guide to host OpenClaw for API integrations.

With your key secured, the last piece is making sure OpenClaw runs on a server with enough resources to handle your workload.

Choosing the right VPS for OpenClaw

OpenClaw 2.0 is built to run continuously, and that shapes what it needs from a server. Its plugin hub, ClawHub skills and MCP connectors need full root access and Docker support to install, update and roll back safely, which shared or restricted hosting typically blocks.

Its automations can resume work missed during a restart, and its messaging channels hold accepted messages through one, but only on a host that stays online and comes back quickly. A laptop that sleeps overnight can’t provide that. A self-managed VPS can.

Bluehost self-managed VPS plans include full root access, Docker support and unmetered bandwidth, and they let you manage AI Gateway credits and keys from the same control panel you use to run OpenClaw.

On our storefront, NVMe 4 is the recommended starting tier and the functional minimum for a working OpenClaw deployment. Here’s how the tiers compare:

Plan	vCPU	RAM	Storage	Best for
NVMe 4	2	4 GB DDR5	100 GB	Starting tier for a single OpenClaw deployment
NVMe 8	4	8 GB DDR5	200 GB	Several channels, multiple Automations and heavier traffic
NVMe 16	8	16 GB DDR5	450 GB	Heavy AI workloads, large teams and growing skills libraries

Every NVMe plan includes unmetered bandwidth. Because your AI Credits wallet sits at the account level, you can also size up later without changing how OpenClaw connects to its models.

Once OpenClaw is running on the right VPS and connected to its models, the next step is connecting it to the channels where your team actually works.

Connect OpenClaw to your messaging channels

Model access is only half of a working OpenClaw setup. The other half is where your agent talks to people, across the messaging platforms your team already uses.

OpenClaw 2.0 can run across Slack, Discord, Telegram, WhatsApp, Signal and Microsoft Teams. Each channel has its own setup, so you connect the ones your team relies on.

Accepted messages persist through a restart, so a pending request is not lost when the server reboots. This channel behavior is separate from how OpenClaw reaches AI models.

If a channel, key or service doesn’t connect as expected during setup, a few common issues are usually the cause.

Troubleshooting common connection issues

A few connection problems show up often during setup and the first runs. Here is how to spot and fix the most common ones:

• API key errors: Check the key is pasted correctly with no extra spaces, and confirm your wallet still has credits.
• Agent task failures: These often mean the server is short on RAM, so size up to a larger VPS tier.
• Gateway service not starting: A port conflict on OpenClaw’s default port 18789 can block startup, so free the port and restart the service.
• Webhook or integration failures: Open the ports the integration needs and check your firewall rules.

Final thoughts

Connecting OpenClaw to AI models no longer has to mean opening an account with every provider. OpenClaw sends each cloud request with an API key, and Bluehost AI Gateway gives you that access through one prepaid wallet, keys that work across supported providers and a single bill. Local models through Ollama skip the key entirely.

From there, setup comes down to a few steps: add AI Credits, create a named key for each app, paste it into OpenClaw and store it securely. Pair that with a self-managed VPS that has root access and Docker, starting at NVMe 4, and your Automations and channels get an always-on home to run from.

When you’re ready, OpenClaw VPS hosting lets you pick your plan and add AI Credits in the same checkout. Already on a Bluehost VPS? Add credits from the AI Tooling tab and connect your first key today.

Once model access stops being the bottleneck, the question shifts from how to connect OpenClaw to what you’ll hand it first. Which repetitive task on your team is ready to become its first Automation?

FAQs