OpenRouter

Create Account and Load Credit

Register for an OpenRouter account and load a small amount of credit to start testing.

Sign up at openrouter.ai, go to the Credits section, load a minimum of 5-10 USD via card to get a working API key.

Start with a small load to test before committing to a larger budget for production.

Retrieve API Key and Configure in Code

Get the API key from the dashboard and configure it in the application using the OpenAI compatible SDK.

Go to Keys > Create Key, copy the key, and configure the base URL `https://openrouter.ai/api/v1` in the OpenAI SDK instead of the default OpenAI URL.

Due to OpenAI SDK compatibility, most existing code calling ChatGPT can be switched to OpenRouter simply by changing the base URL and key.

Select Specific Model to Call

Specify the correct model name according to OpenRouter's format when calling the API.

Pass the model parameter in the format 'anthropic/claude-sonnet-4' or 'openai/gpt-4o' in the request body according to the model list on openrouter.ai/models.

Check the pricing and average latency of each model directly on the models page before officially choosing.

Configure Fallback List

Set up a list of backup models to automatically switch when the main model does not respond.

Pass an additional parameter `models: ["openai/gpt-4o", "anthropic/claude-sonnet-4"]` so that OpenRouter tries them sequentially if the first model fails.

Order the fallback list based on priority in terms of both quality and cost, not just on stability.

Monitor Usage and Set Budget Limits

Monitor spending and set limits to avoid exceeding the budget out of control.

Go to the Activity section on the dashboard to view detailed usage by model, set a maximum spending limit in Settings.

Set up an email alert when usage reaches 80% of the budget to allow timely adjustments before running out of credit.