How to Connect Bluehost AI Credits to OpenClaw

This article covers connecting an AI Credits API key to OpenClaw so its agents can run on your AI Credits balance instead of a separate LLM provider account. It assumes OpenClaw is already installed on your Self-Managed VPS or VDS. Use Section 1 if you haven't configured a model provider yet, or Section 2 if OpenClaw is already set up with a different provider and you want to add or switch to AI Credits.

Section 1: First-Time Setup (No Provider Configured Yet)

OpenClaw's first run doesn't use a traditional menu wizard — it hands you to a conversational setup agent (named "Crestodian" in current versions) that asks what you want in plain language instead of numbered menus. The steps below walk through that flow with the actual screens you'll see.

1. Connect to your server over SSH, then run:
   Copy

   openclaw

   On first run, the setup agent greets you and offers to configure the basics (workspace, gateway) before asking about a model provider:

   Hi, I'm Crestodian — let's hatch your agent.
   No menus here: tell me what you want and I'll do the configuring. I looked around this machine:
   - AI: nothing detected yet (no Claude Code or Codex login, no OPENAI_API_KEY/ANTHROPIC_API_KEY). I'll
     set up the basics first, then ask whether you want to configure a model provider with masked
     credential prompts.
   - Workspace: ~/.openclaw/workspace
   - Gateway: runs locally, private to this machine (token auth).
   Say yes and I'll set all of that up now.

   Reply:

   Copy

   yes
2. It sets up the workspace and Gateway service (this includes enabling systemd lingering so the Gateway keeps running after you log out):

   Workspace OK: ~/.openclaw/workspace
   Sessions OK: ~/.openclaw/agents/main/sessions
   Systemd: Enabled systemd lingering for root.
   Gateway service installed.
   Updated /root/.openclaw/openclaw.json
   Workspace: ~/.openclaw/workspace
   Gateway: running at ws://127.0.0.1:18789
   Default model: not configured yet

   Then it asks:

   No usable model provider is configured, so the agent cannot answer yet.
   Configure a model provider now? Say yes or no.

   Reply yes.

3. You'll see a short list of named providers plus a More… option:

   ◆  Model/auth provider
   │  ● OpenAI (ChatGPT/Codex sign-in or API key)
   │  ○ Anthropic
   │  ○ xAI (Grok)
   │  ○ Google
   │  ○ OpenRouter
   │  ○ Meta
   │  ○ More…
   │  ○ Skip for now
   │  ↑/↓ to navigate • Enter: confirm

   AI Credits isn't one of the named providers, so select More…. This opens a searchable, expanded list:

   │  Search: _
   │  ● OpenAI (ChatGPT/Codex sign-in or API key)
   │  ○ Anthropic
   │  ○ xAI (Grok)
   │  ○ Google
   │  ○ OpenRouter
   │  ○ Arcee AI
   │  ○ BytePlus
   │  ○ Cerebras
   │  ○ Chutes
   │  ○ ClawRouter
   │  ○ Cloudflare AI Gateway
   │  ○ Codex
   │  ○ Cohere
   │  ○ Copilot
   │  ○ Custom Provider
   │  ○ DeepInfra
   │  ○ DeepSeek
   │  ○ Featherless AI
   │  ○ Fireworks
   │  ○ GMI Cloud
   │  ○ Google Vertex
   │  ○ Groq
   │  ○ Hugging Face
   │  ○ Kilo Gateway
   │  ○ LiteLLM
   │  ○ LM Studio
   │  ...
   │  ↑/↓ to select • Enter: confirm • Type: to search

   Find and select Custom Provider.

4. At the base URL prompt, enter your gateway endpoint:

   ◇  Model/auth provider
   │  Custom Provider
   │
   ◆  API Base URL
   │  https://api.example.com/v1

   Copy

   https://gateway.ai.bluehost.com
5. Next, it asks how you want to provide the API key:

   │  https://gateway.ai.bluehost.com
   │
   ◆  How do you want to provide this API key?
   │  ● Paste API key now (Stores the key directly in OpenClaw config)
   │  ○ Use external secret provider
   │  ↑/↓ to navigate • Enter: confirm

   Select Paste API key now, then enter your AI Credits API key at the following prompt:

   │  Paste API key now
   │
   ◆  API Key (leave blank if not required)
   │  _

6. At the compatibility prompt, select the first option:

   ◆  Endpoint compatibility
   │  ● OpenAI-compatible (Uses /chat/completions)
   │  ○ OpenAI Responses-compatible
   │  ○ Anthropic-compatible
   │  ○ Unknown (detect automatically)
   │  ↑/↓ to navigate • Enter: confirm

   OpenAI-compatible (Uses /chat/completions) matches how the AI Credits gateway is built. Don't select "OpenAI Responses-compatible" — that's a different, newer format the gateway doesn't use.

7. At the Model ID prompt, enter the exact model ID you want to use as your default:

   │
   ◆  Model ID
   │  e.g.

   Copy

   claude-haiku-4-5

   See AI Models and Credit Pricing for the full list of supported model IDs.

8. If everything's entered correctly, you'll see Verification successful, then prompts for an Endpoint ID and an optional Model alias:

   ◇  Model ID
   │  claude-haiku-4-5
   │
   ◇  Verification successful.
   │
   ◆  Endpoint ID
   │  custom-gateway-ai-bluehost-com█

   The Endpoint ID is just an internal label for this provider — the auto-suggested value is fine to keep. Press Enter to accept it, then you'll see an optional Model alias prompt:

   ◇  Endpoint ID
   │  custom-gateway-ai-bluehost-com
   │
   ◆  Model alias (optional)
   │  e.g. local, ollama

   This is a short nickname for the model; safe to leave blank and press Enter to skip.

Once setup finishes, you'll land back with the setup agent, now showing your connected provider and model:

Hi, I'm Crestodian.
- Start me when setup, config, Gateway, model choice, or agent routing feels off.
- Using: custom-gateway-ai-bluehost-com/claude-haiku-4-5 — just tell me what you want.
- Config: valid. Default agent: main.
- Gateway: reachable at ws://127.0.0.1:18789.
Everything basic is reachable. Use talk to agent when you want the normal agent.

Section 2: Adding AI Credits to an Existing OpenClaw Install

If OpenClaw is already configured with a different provider and you want to add AI Credits (or switch to it) without repeating full setup, edit OpenClaw's configuration file directly.

1. Connect to your server over SSH, then open OpenClaw's configuration file:
   Copy

   nano ~/.openclaw/openclaw.json
2. Add an entry for AI Credits under models.providers, using your gateway endpoint and API key:
   Copy

   { "models": { "providers": { "bluehost-ai-credits": { "baseUrl": "https://gateway.ai.bluehost.com", "apiKey": "YOUR_AI_CREDITS_API_KEY", "api": "openai-completions", "models": [ { "id": "MODEL_ID", "name": "Display Name" } ] } } }, "agents": { "defaults": { "model": { "primary": "bluehost-ai-credits/MODEL_ID" } } } }
3. Save and close the file.

Adding the new provider this way doesn't remove your existing provider configuration — only the agents.defaults.model.primary value determines which one OpenClaw actually uses by default.

Testing the Connection

If you just finished Section 1, the setup agent already confirmed the connection with Verification successful during setup. To also confirm the model responds correctly, hand off to the agent and send a message:

Example Output:

Handing you over to agent main now. It's using Claude Haiku 4.5 via your custom gateway. Go
ahead—what do you need?

Hey. I just came online. Who am I? Who are you?
Let me figure this out with you.

First things first:
1. What's your name? (What do I call you?)
2. What should I be? (AI assistant is solid, but I can be weirder if you want — familiar, ghost in
   the machine, something else entirely?)
3. How should I come across? (Sharp? Warm? Chaotic? Chill?)
4. Pick me an emoji. (Every good assistant needs a signature.)

The agent's first response is its own onboarding conversation (asking what to call it, its personality, and so on) — that's expected and unrelated to AI Credits. What matters is that it responded at all, and that the status line at the bottom of the screen shows your custom provider and model (for example, custom-gateway-ai-bluehost-com/claude-haiku-4-5), confirming the reply came through your AI Credits gateway.

Summary

On a fresh OpenClaw install, run openclaw and follow its conversational setup agent through workspace/Gateway creation, then choose a model provider. Since AI Credits isn't one of the short-listed named providers, select More… to find Custom Provider, then enter your gateway endpoint, API key, OpenAI-compatible (Uses /chat/completions) as the compatibility type, and your chosen model ID. A Verification successful message confirms the connection immediately during setup. For an existing install with a different provider already configured, try asking the setup agent conversationally first, or edit ~/.openclaw/openclaw.json directly as a fallback. Either way, confirm everything works end-to-end with talk to agent and a test message.