MetisRouter Troubleshooting Guides

Set the API base URL to https://api.metisrouter.com/v1, paste a MetisRouter API key, choose a published chat model ID, then send a short test prompt.

Choose OpenAI Compatible, set MetisRouter as the base URL, use Bearer API key authentication, and copy the exact model ID from the catalog.

Use an HTTP Request or OpenAI Chat Model node with Bearer auth, set max_tokens, and log the request id for support.

Use https://api.metisrouter.com/v1 as the API base URL, confirm the API key can call /v1/models, then refresh the provider model list.

Use Anthropic Direct only for models tagged Anthropic Direct; otherwise use an OpenAI-compatible coding model through the documented path.

Open /models, copy the exact API model ID from the model drawer, and make sure the model supports the OpenAI-compatible text endpoint.

Set stream=true only for text endpoints that support streaming, keep max_tokens bounded, and verify the request in Usage Logs.

Choose a model tagged for tool calling or structured output, pass tools in the OpenAI-compatible request body, and keep tool output short enough for context.

Use smaller bounded requests, model-specific async task patterns for long generation jobs, and retry only after checking Usage Logs.

Keep at least 5 USD available balance and set max_tokens, max_completion_tokens, or max_output_tokens before sending low-balance requests.

Create a MetisRouter API key, paste it into Cursor's OpenAI-compatible provider settings, keep the base URL on /v1, and test with a published chat model ID.

Store the MetisRouter key in the environment variable expected by the client, set the OpenAI-compatible base URL, and run one bounded test request.

Use streaming when supported, bound output with max_tokens or max_output_tokens, split large repository tasks, and inspect Usage Logs before retrying.

Choose OpenAI Compatible in Cline, set the MetisRouter base URL, paste the API key, copy an exact Qwen coding model ID from /models, and verify Usage Logs.