OpenRouter

OpenRouter is a service that acts as an intermediary between your application and language model providers. Instead of maintaining separate accounts and API keys for OpenAI, Anthropic, Google, and DeepSeek, you get one key and one balance, while selecting the model by specifying it in the request: openai/..., anthropic/..., deepseek/.... According to the service, more than 600 models from over 80 providers are available through it.

The OpenRouter API is compatible with OpenAI: in any application that supports OpenAI, simply replace the endpoint with https://openrouter.ai/api/v1 and enter your OpenRouter key. That is why the service is supported by n8n, Cursor, Cline, SillyTavern, LibreChat, and many other clients.

Pricing and fees

• There is no markup on models: tokens cost the same as they do from the provider itself.
• The fee is charged when adding funds: 5.5% on the Standard plan (pay-as-you-go, with no subscription fee or minimum commitment). The Business plan, with an 8% fee, adds routing within the EU or US and DPA signing; Enterprise offers custom terms, SSO, and SLA.
• Payment methods: bank cards, AliPay, and USDC cryptocurrency; automatic top-ups are also available. Under the service’s terms, purchased credits may expire one year after purchase.
• Bring Your Own Keys (BYOK): you can connect your own OpenAI, Anthropic, and other provider keys. The first $25,000 of monthly spending (at OpenRouter pricing) is commission-free; beyond that, the fee is 5%.

Free models

Some models have a free version with the :free suffix; more than 25 such models are available on the free plan. Limits:

Failed requests also count toward the daily limit, and free models are disabled when the balance is negative. OpenRouter does not impose its own plan-specific limits on paid models; restrictions are set by providers.

What OpenRouter can do

• Fallback providers: if a provider returns an error, the request is automatically sent to the next one; tokens are not charged for failed attempts.
• Automatic model selection: openrouter/auto selects a model based on the request, while openrouter/free selects a free model.
• Routing options: the :nitro suffix selects the fastest provider, :floor selects the cheapest one, and :exacto selects providers with reliable tool calling.
• Privacy: by default, requests and responses are not stored. In the settings, you can disable providers that store data or use it for training, and enable routing only through providers with Zero Data Retention.
• Prompt caching, budgets, and separate keys for projects are available on paid plans.

Who is it for?

• Those comparing models: switch between Claude, GPT, Gemini, and DeepSeek with a single line, without creating new accounts.
• Agent and automation developers: use one key for all models, with fallback providers in case of outages.
• Those who find it inconvenient to pay for multiple services: use one balance instead of several API subscriptions.

If the lowest possible price for one specific model is important, it is worth getting it directly from the provider: the 5.5% top-up fee is the price of convenience.

How to connect OpenRouter to n8n — for agents, workflows, and the n8n Assistant: Get started with n8n.

Do I need a subscription to use OpenRouter?

No. You pay only for the tokens you use from a pre-funded balance. Free models can be used without adding funds, with a limit of 50 requests per day.

Is OpenRouter more expensive than using a provider’s API directly?

Token prices are the same; the difference is the 5.5% top-up fee on the Standard plan.

Does OpenRouter store my requests?

Not by default. However, the request is still sent to the model provider, which has its own data retention policy. Providers that store data can be disabled in the privacy settings.