AirAPI Provider for OpenClaw

Use AirAPI models from OpenClaw with a single provider setup.

This plugin connects OpenClaw to the shared AirAPI-compatible endpoint at:

• https://ap-1.aieev.cloud:2007/ai/api/v1

It reads model metadata from:

• https://ap-1.aieev.cloud:2007/ai/api/v1/models

What this plugin does

• Adds one OpenClaw provider: airapi
• Prompts for AIRAPI_API_KEY during setup
• Reads compatible chat-model metadata from AirAPI /models
• Preserves exact server-returned model ids
• Uses per-model endpoint_url values when the API advertises them
• Falls back to known per-model base URLs for the currently supported Qwen models when endpoint_url is missing
• Supports model refs such as:

Install

Once published to ClawHub:

• openclaw plugins install clawhub:openclaw-airapi-provider

For a local checkout:

• openclaw plugins install /path/to/openclaw-airapi-provider

Authenticate

Run:

• openclaw models auth login --provider airapi

Or use onboarding:

• openclaw onboard
• choose AirAPI
• enter your AirAPI API key

You can also provide the key with:

• --airapi-api-key <YOUR_AIRAPI_API_KEY>

Environment variable:

• AIRAPI_API_KEY

Model metadata and supported models

This plugin fetches model metadata from:

• GET https://ap-1.aieev.cloud:2007/ai/api/v1/models

Only compatible chat models are exposed to OpenClaw. Non-chat models such as TTS-only models are ignored.

Current intended AirAPI LLM support in this plugin:

• qwen/qwen3.5-9b
• qwen/qwen3.5-35b-a3b

The plugin still reads /models for metadata and exact model ids, but it also carries explicit fallback routing for the supported Qwen models so they keep their intended base URLs even if endpoint_url is absent from discovery.

Base URL behavior

Default provider base URL:

• https://ap-1.aieev.cloud:2007/ai/api/v1

Default discovery URL:

• https://ap-1.aieev.cloud:2007/ai/api/v1/models

If the AirAPI /models response includes a model-specific endpoint_url, this plugin uses that per model.

This means one provider id (airapi) can expose multiple LLM models with different baseUrl values. The provider-level baseUrl is the shared fallback/discovery URL, while each model may override it with its own serving endpoint.

Local validation completed on this machine:

• OpenClaw loaded the plugin under the single provider id airapi
• OpenClaw cached provider metadata with provider base URL https://ap-1.aieev.cloud:2007/ai/api/v1
• OpenClaw cached model entries for the same provider with distinct per-model base URLs, including:

The provider base URL and default recommended model are fixed in the plugin source to follow the documented OpenClaw provider-plugin pattern and avoid raw environment-variable reads in plugin runtime code.

Usage

Example model refs:

• airapi/qwen/qwen3.5-9b
• airapi/qwen/qwen3.5-35b-a3b

Check configured model state:

• openclaw models status --plain

Troubleshooting

• AirAPI does not appear in onboarding

• API key prompt works but inference returns 404 model not found

• No models appear

• A different provider warns during onboarding (for example vLLM)

Development

Requirements:

• Node.js 20+
• pnpm

Install:

• pnpm install

Test:

• pnpm test

Watch tests:

• pnpm test:watch

Current implementation notes

• Provider id: airapi
• Plugin id: openclaw-airapi-provider
• Package name: @aieev/openclaw-airapi-provider
• API transport: openai-completions