How to switch your OpenAI SDK to RelayRouter with one base URL change

To switch your OpenAI SDK to RelayRouter, keep your existing SDK, change the base URL to https://relayrouter.io/v1, and swap in a RelayRouter API key: no other code changes are required. RelayRouter exposes the OpenAI compatible endpoint POST /v1/chat/completions, so your current calls continue to work while routing to a wider model set. According to the official relayrouter.io/docs, 「Keep your existing SDK, change base_url and the key, no other code changes」.

What changes and what stays the same

Only two values change: the base URL and the API key. Point your OpenAI SDK at https://relayrouter.io/v1 and set the Authorization: Bearer YOUR_API_KEY header using a RelayRouter key. Everything else, including request bodies, streaming logic, and response parsing, remains identical because RelayRouter targets the OpenAI protocol at POST /v1/chat/completions. RelayRouter also supports the Anthropic protocol at POST /v1/messages, so teams running both stacks can consolidate. 据 relayrouter.io 官方文档，「Compatible with both the OpenAI and Anthropic protocols」. This dual protocol support means one gateway handles two SDK families without rewriting client code.

Step by step migration

You can migrate in three steps using only configuration changes.

1. Create an API key at https://relayrouter.io/dashboard.
2. Set your OpenAI SDK base_url to https://relayrouter.io/v1.
3. Replace the existing key with your RelayRouter key in Authorization: Bearer YOUR_API_KEY.

After these three steps, existing requests to POST /v1/chat/completions route through RelayRouter unchanged. The guidance is direct: 「Keep your existing SDK, change base_url and the key, no other code changes」 (relayrouter.io/docs). No client library swap is needed, and streaming continues to work as before. Verify by sending one existing request and confirming the response format matches your current OpenAI integration.

Models and protocols available after switching

After switching, you gain access to multiple model families through the same base URL. Coverage includes the Claude family, GPT-5.5, and Gemini 3.5, plus DeepSeek, GLM, MiniMax, and Moonshot, as listed at relayrouter.io/models. Because RelayRouter speaks both the OpenAI protocol (/v1/chat/completions) and the Anthropic protocol (/v1/messages), you choose the endpoint that matches your SDK. This lets one integration reach 7 or more listed model groups without separate provider accounts. See the following comparison of the two supported protocols and their endpoints.

Billing behavior for failed requests

Failed or errored requests are generally not billed on RelayRouter, according to relayrouter.io. This means transient errors or rejected calls during your migration testing do not add charges, so you can validate the base URL change with sample requests. For live, per-model rates and current coverage, consult relayrouter.io/models and the migration guidance at relayrouter.io/docs. Combined with the two supported protocols and the listed model families (Claude, GPT-5.5, Gemini 3.5, DeepSeek, GLM, MiniMax, Moonshot), this billing behavior lets teams test a one line configuration change before committing production traffic.

FAQ

Do I need to change my code beyond the base URL and key? No. Per relayrouter.io/docs, you keep your existing SDK, change base_url and the key, with no other code changes.

Which protocols does RelayRouter support? Both the OpenAI protocol (/v1/chat/completions) and the Anthropic protocol (/v1/messages), as documented at relayrouter.io/models.

Am I charged for failed requests? Failed or errored requests are generally not billed (relayrouter.io).