OpenClaw

OpenClaw can use HaiToken through an OpenAI-compatible provider in openclaw.json. The stable path is to register HaiToken under models.providers, then point the default agent model to that provider.

Protocol Note

OpenClaw’s custom provider integration uses the openai-completions API type, so it only supports the OpenAI Chat Completions protocol; it does not support the Anthropic Messages protocol. This guide therefore only covers the OpenAI mode, with the Base URL fixed to https://api.haitoken.ai/v1.

What you need

A HaiToken API key
At least one available model ID, such as gpt-5.4
HaiToken OpenAI-compatible base URL: https://api.haitoken.ai/v1

If you see ~, environment variables, or user-directory placeholders in OpenClaw examples, read them as “your current user home directory.” On Windows this is usually %USERPROFILE%, on macOS usually /Users/<your-name>, and on Linux usually /home/<your-name>.

Setup Steps

Step 1: Edit `openclaw.json`

Add a HaiToken provider under models.providers. Example:

{
  models: {
    mode: "merge",
    providers: {
      haitoken: {
        baseUrl: "https://api.haitoken.ai/v1",
        apiKey: "${HAITOKEN_API_KEY}",
        api: "openai-completions",
        models: [
          { id: "gpt-5.4", name: "GPT-5.4" },
          { id: "gpt-5.4", name: "GPT-5.4" },
          { id: "gpt-4o", name: "GPT-4o" }
        ]
      }
    }
  },
  agents: {
    defaults: {
      model: {
        primary: "haitoken/gpt-5.4",
        fallbacks: ["haitoken/gpt-5.4"]
      },
      models: {
        "haitoken/gpt-5.4": { alias: "flash" }
      }
    }
  }
}

Step 2: Set the environment variable

Windows PowerShell:

$env:HAITOKEN_API_KEY = "YOUR_API_KEY"

macOS / Linux:

export HAITOKEN_API_KEY="YOUR_API_KEY"

Step 3: Reload OpenClaw

After saving the file, first check whether OpenClaw picks up the change automatically. If the new model does not appear, restart the gateway or reopen OpenClaw and test again.

Step 4: Pick the HaiToken-backed model

Use the primary model or one of the aliases you configured and send a simple prompt.

Notes

Item	Value
Provider base URL	`https://api.haitoken.ai/v1`
API type	`openai-completions`
Chat endpoint	`POST /v1/chat/completions`
Auth header	`Authorization: Bearer YOUR_API_KEY`

Do not write OpenClaw setup as an install or onboarding guide here. The key part is the provider configuration and model mapping.

Model discovery through GET /v1/models can help when debugging, but you should not describe it as a guaranteed mandatory step for every OpenClaw flow.

Verification

Open OpenClaw.
Select the HaiToken-backed model or alias.
Send a simple prompt.
If the response returns normally, the integration is working.

FAQ

Question	Fix
`401` or authentication failed	Make sure `HAITOKEN_API_KEY` is set in the same environment where OpenClaw is running.
`model not found`	Make sure the model IDs in `models.providers.haitoken.models` exactly match the IDs exposed by HaiToken.
The configuration was saved but the new model does not show up	OpenClaw may not reload immediately in every setup. Reopen it or restart the related gateway process and test again.
Do I always need to restart?	No. Check whether the config has been reloaded automatically first. Restart only if the updated provider or models are still missing.

Quick start

API documentation

AI tool integration

Other

Protocol Note

What you need

Setup Steps

Step 1: Edit `openclaw.json`

Step 2: Set the environment variable

Step 3: Reload OpenClaw

Step 4: Pick the HaiToken-backed model

Notes

Verification

FAQ

Next

​Protocol Note

​What you need

​Setup Steps

​Step 1: Edit openclaw.json

​Step 2: Set the environment variable

​Step 3: Reload OpenClaw

​Step 4: Pick the HaiToken-backed model

​Notes

​Verification

​FAQ

​Next

Protocol Note

What you need

Setup Steps

Step 1: Edit `openclaw.json`

Step 2: Set the environment variable

Step 3: Reload OpenClaw

Step 4: Pick the HaiToken-backed model

Notes

Verification

FAQ

Next