Client tools
Connect chat applications and coding agents to the VNAI Pro API
These guides use an API key created on your environment's website, with the endpoint and models available to that key.
Which tools can connect?
| Tool | Connection method | Constraint |
|---|---|---|
| OpenClaw | Custom provider | Match the model's API protocol |
| Hermes Agent | Custom endpoint | Configure with hermes model |
| OpenCode | Custom provider | Distinguish Chat Completions from Responses |
| Cursor | OpenAI key + Override Base URL | Conditional support; excludes Tab completion |
| Qwen Code | Custom Provider / settings.json | Qwen models are not required |
| Chatbox | Custom provider | API Host and API Path are separate |
| Cline | OpenAI Compatible | Agent tasks require model tool calling |
| DeepSeek Harness | Custom model API | Three protocols; currently a developer preview |
| 9Router | OpenAI-compatible provider | Add the VNAI Pro API as an upstream provider |
Verification scope: official documentation/source reviewed on September 27, 2026, against the gateway's API protocols. These are configuration guides, not live inference certification for every tool/model/group. Verify model access and tool calling with your own key.
1. Prepare a key and model
- Follow the API key guide at https://api.vnaipro.app/keys; check your balance at https://api.vnaipro.app/wallet and the key's remaining quota.
- Choose the Auto group to share one key across Codex, Claude, and other supported models, or a dedicated group containing your desired models.
- Open Check models, enter the key, and copy the exact model ID.
- Replace
YOUR_API_KEYand everyYOUR_MODEL_IDin the guide.
The /v1/models catalog does not by itself prove support for every protocol, images, reasoning, or tool calling.
2. Choose the URL and protocol
| Purpose | URL for this environment |
|---|---|
| Gateway root | https://api.vnaipro.app |
| OpenAI Chat Completions / Responses base URL | https://api.vnaipro.app/v1 |
| Model catalog | https://api.vnaipro.app/v1/models |
| Chat Completions request | https://api.vnaipro.app/v1/chat/completions |
| Responses request | https://api.vnaipro.app/v1/responses |
| Anthropic Messages request | https://api.vnaipro.app/v1/messages |
Most Base URL fields take the URL ending in /v1, not the full request path.
Anthropic adapters differ; follow the individual guide. Do not use the docs domain as an API endpoint.
Paste the raw key into API key, without a Bearer prefix; the client adds authentication.
Keep keys in personal configuration and do not commit files containing keys.
3. Verify the connection
- Start a new conversation and select the configured provider and model.
- Send “Reply only OK” to check authentication and responses.
- For coding agents, use a test directory and ask for a file listing or a file read. Keep tool approval enabled.
- Check usage history on the website to confirm requests reached your gateway.
A text response alone does not validate an agent: the model and upstream channel also need streaming and tool/function calling. Search, image, speech, and embedding services may need separate endpoints or keys.
Troubleshooting
| Error | Action |
|---|---|
401/403 | Check the key, environment, group access, balance, and key restrictions |
404 or HTML response | Check the API domain and protocol; avoid duplicated /v1/v1 |
| Model not found | Use an ID from Check models, not a display name |
| Chat works but tools fail | Verify tool calling on the model/channel and choose the correct adapter |
400 for unsupported parameters | Read the error; remove unsupported reasoning, image, or request options |
429 | Check quota and reduce concurrent requests/agents |