OpenAI Compatible Base URL Checker
This is an OpenAI-compatible API tester: check repeated /v1, complete endpoint paths, query parameters, and client URL-joining rules. All analysis stays in the current browser.
Link base addresses for the OpenAI SDK and Claude Code
Use https://www.aifast.link/v1 in the OpenAI SDK, with /v1/models as the final model-list path. Set ANTHROPIC_BASE_URL to https://www.aifast.link for Claude Code; Messages requests reach /v1/messages. This tool's OpenAI path preview does not validate the Messages protocol.
A correct path does not establish plan coverage. Confirm the Link key, key group and plan scope before sending a minimal request and checking its usage log. This checker does not call the API or incur model charges.
Enter a Base URL
Enter your current Base URL to see the suggested setting and final endpoint URLs.
Diagnosis details
Three Base URL input and output examples
These examples check URL strings only. They do not call a model, validate a key or confirm endpoint availability. The suggested value depends on whether the client appends /v1; always compare it with the final request URL in client logs.
| Configuration and input | Appends /v1 | Suggested Base URL | Resulting endpoint URL |
|---|---|---|---|
Duplicated /v1 and a full chat endpointhttps://www.aifast.link/v1/v1/chat/completions | No | https://www.aifast.link/v1 | https://www.aifast.link/v1/chat/completions |
A Responses endpoint entered as the base URLhttps://www.aifast.link/v1/responses | No | https://www.aifast.link/v1 | https://www.aifast.link/v1/responses |
The client appends another /v1https://www.aifast.link/v1 | Yes | https://www.aifast.link | https://www.aifast.link/v1/chat/completions |
If the path is correct but the API returns 401, check authentication. For a model-not-found error, check the model ID and group permissions. Adding funds or repeatedly replacing keys will not fix URL-joining errors.
Troubleshoot 401 and authentication · Check model IDs and permissions
Content updated:
How do you check an OpenAI-compatible API URL?
Paste the current configuration into the tool first, then compare the final URL with client logs. Whether the field is called API Host, Endpoint, or Base URL matters less than where the client actually sent the request.
- 01
First check if there are duplicate paths
/v1/v1、/chat/completions/chat/completionsindicate that the path was appended more than once. - 02
Check the response type
An HTML login page usually means the request did not reach an API route. For a JSON error response, check authentication, the model ID and request parameters next.
- 03
Then send a minimal request
List the available models, then send a short text request. Test SSE streaming, tool calls and multimodal input separately after the basic request succeeds.
Base URL FAQ
This tool checks how URLs are assembled. Authentication and model capabilities require separate live requests.
Should the Base URL contain /v1?
It depends on whether the client appends /v1. An OpenAI SDK custom base_url usually includes /v1. If your client explicitly appends it, use the host or parent path instead. Confirm the final URL in request logs rather than relying on the field label.
Why does /v1/v1/chat/completions appear?
The configuration value already contains /v1, and the client appends it automatically. One of the layers should be removed based on client rules and the final request address should be rechecked.
Will this tool request the API gateway I filled in?
No. The checking process only analyzes the URL string in the current browser, does not send network requests, and does not require an API Key. To verify real API responses, use model checking tools.
Why does the checker warn about a full chat/completions URL?
Most clients expect a base path and append models, chat/completions or responses themselves. Entering a full endpoint URL as the Base URL can duplicate the endpoint path.
How do you check an OpenAI-compatible API URL?
Paste the Base URL currently used by the client into the checker. Preview the final models, chat/completions, and responses URLs, then use client logs to confirm whether /v1 was appended twice. This tool analyzes only the URL string; it does not verify your API key or a live response.
Continue reading Model not found and /v1/v1 troubleshooting、OpenAI Compatible quick access, or view OpenAI API Reference .