FAQ
Find concise answers about AGTCloud API integration, authentication, model selection, and common errors.
This page answers the questions developers most often have when integrating with AGTCloud for the first time. For endpoint fields, model capabilities, and platform-specific error codes, continue to rely on the corresponding documentation and actual responses.
Follow Quickstart to sign in, create an API key, select a model, and send a minimal request. For the first test, start with an OpenAI Chat-compatible request. After the basic request succeeds, add streaming output, tool calls, or multimodal input.
Open API Keys to create, disable, or delete tokens. Use different keys for different environments. Promptly disable or delete keys that are no longer used or may have been exposed.
The recommended Base URL is https://api.agtcloud.cn/v1. When using the OpenAI SDK, set it as base_url or baseURL. OpenAI-compatible endpoints usually authenticate with Authorization: Bearer your-api-key. For other protocols, follow the authentication header requirements in the corresponding API Reference. See API keys and authentication for details.
This is not recommended. Store long-lived API keys on a trusted server, in server-side environment variables, or in a secret manager. Browser and mobile applications should call your own backend, which then calls AGTCloud, to reduce the risk of users extracting and abusing the key.
A common cause is that the Base URL already contains /v1 and the code adds another /v1 before the endpoint. Use https://api.agtcloud.cn/v1 as the SDK Base URL, then let the SDK append endpoints such as /chat/completions. When troubleshooting, inspect the complete URL that was actually sent.
Open the Model marketplace to view the models and model IDs currently displayed. Model visibility can change with platform configuration and account permissions. Copy the ID shown on the current page and rely on the actual request result.
If you do not need to preserve provider-specific protocol semantics, start with OpenAI Chat Completions because its SDK and integration pattern are widely used. For an existing Anthropic, Gemini, or Responses application, keep the current SDK and message structure first, then confirm how the target model is supported.
For 401, check the authentication headers required by the protocol and the API key. For 403, check the account, key, and model permissions. For 404, check the Base URL, endpoint, model ID, or task ID. Do not retry these errors indefinitely; fix the identified cause first. See Errors and troubleshooting for more steps.
Check the response headers for x-request-id, request-id, or x-oneapi-request-id, and save the value that is actually present. Some error bodies can also contain a Request ID. If none is available, still record the request time, Method, Path, model, and HTTP status. See Request ID for code examples.
Immediately open API Keys to disable or delete the old key. Then create a new key and update the services that use it. Check repositories, logs, screenshots, and chats for remaining copies. If the key was committed to Git, deleting it from the current file is not enough; follow your team's security process for repository history.
Prepare the request time and time zone, Method, Path, model ID, HTTP status, redacted error information, SDK and runtime versions, and minimal reproduction steps. Include the Request ID when the response contains one. Do not submit complete API keys, authentication headers, cookies, customer prompts, or private files.
Still need help?
Complete the basic checks in Errors and troubleshooting, then use Request ID to prepare redacted diagnostic information. Do not submit complete API keys, authentication headers, cookies, or sensitive customer data to unknown destinations.