NEW

Jev is now available

Model prices

Troubleshoot an ApiClo API connection

Start with the HTTP status and response body. A 401 usually concerns authentication; 404 can mean the wrong endpoint or model; 403 requires checking access and the requested model; 429 calls for checking limits or temporary capacity. Keep the error body and request ID when contacting support.

Before you start

  • Record the client version, requested model, endpoint and approximate request time.
  • Keep credentials out of screenshots, logs shared publicly and support messages.
Create an API key

Steps

  1. Check address and key

    Use https://apiclo.com/v1 for an OpenAI-compatible Base URL, or https://apiclo.com for Claude Code's ANTHROPIC_BASE_URL. Test GET /v1/models with your key to isolate authentication from generation.

  2. Inspect the failed request

    Compare the response body with ApiClo Logs. Verify the model ID exactly as listed; a successful models request does not prove model generation or tool compatibility.

  3. Retry according to the error

    Correct authentication or path errors before retrying. For 429, honor Retry-After when present and use increasing delays. If a generation timed out after starting, inspect its status before resending it.

Examples

macOS / Linux

curl https://apiclo.com/v1/models \
  -H "Authorization: Bearer YOUR_APICLO_KEY"

Check the result

After correcting the cause, send one short generation and verify the model, response and usage record. If it still fails, provide support with the sanitized error and request ID.

Troubleshooting

  • 401: check key validity, whitespace and the header required by the chosen endpoint.
  • 403/404: check the response body, account access, exact model ID and duplicated /v1 segments.
  • 429: rapid repeated retries can prolong congestion; keep the same key and model while backing off.
  • Unexpected charge or timeout: check Logs before repeating a potentially completed generation.

Compatibility and billing

HTTP status alone cannot identify every upstream or client issue. A model-list response validates that request only. Never send your full API key to support.

All documentation