API Overview

RouteAPI provides enterprise-oriented unified AI API access, aggregating model capabilities from OpenAI, Claude, Gemini, Azure, AWS Bedrock, and more into a single stable, observable, and measurable interface system. Business systems only need to integrate with RouteAPI to call different model services under unified authentication, unified model IDs, and unified logging.

Get started in three steps

If you’re already using any OpenAI-compatible SDK, you can get it running by changing just two settings:

curl https://api.routeapi.ai/v1/chat/completions \
  -H "Authorization: Bearer $ROUTEAPI_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.5",
    "messages": [{ "role": "user", "content": "你好" }]
  }'

1. Change the Base URL to https://api.routeapi.ai/v1.
2. Change the API Key to your RouteAPI Token (create one on the API Keys page in the console).
3. Change model to a model ID available on your account. Look it up with GET /v1/models.

For the complete migration steps for an existing OpenAI application, see Migrating from OpenAI.

Supported protocol entry points

RouteAPI supports three mainstream protocols at once: OpenAI-compatible, Claude Messages, and Google Gemini. You can keep using your existing SDK or client and just switch the Base URL and API Key to RouteAPI.

Protocol	Typical endpoints	Suitable scenarios
OpenAI compatible	/v1/chat/completions, /v1/responses, /v1/embeddings	OpenAI SDK, Cursor, OpenCode, LangChain, LiteLLM, and other compatible clients
Claude Messages	/v1/messages	Claude Code, Anthropic SDK, and clients using Claude’s native message format
Google Gemini	/v1beta/models/{model}:generateContent	Google GenAI SDK, Gemini REST clients

Requests entering through different protocols are adapted internally by RouteAPI as needed. On the business side, it’s best to pick whichever protocol your current client natively supports — a native protocol skips conversion entirely and behaves the most predictably. For the mapping rules and known limitations of cross-protocol calls (for example, calling a Claude model using the OpenAI format), see Protocol conversion.

Base URLs

The OpenAI-compatible and Claude Messages protocols use this by default:

https://api.routeapi.ai/v1

The Google Gemini protocol uses this by default:

https://api.routeapi.ai/v1beta

Common request headers

Authorization: Bearer sk-your-routeapi-token
Content-Type: application/json

All protocols use the same type of RouteAPI Token. Keep your Token on the server side — never expose it in a browser, mobile app, or public repository. For details on key permissions, quotas, and authentication errors, see Authentication.

The Claude Messages protocol also accepts the x-api-key header that the Anthropic SDK typically uses, so if you’re using the Anthropic SDK you usually only need to change the Base URL.

API and capability index

Pick the entry point based on what you’re trying to do:

What I want to do	Which endpoint	Docs
Standard multi-turn chat	/v1/chat/completions	Chat Completions
Coding agents, new-protocol clients	/v1/responses	Responses
Vector search, semantic search, RAG	/v1/embeddings	Embeddings
Look up available models and capabilities	/v1/models	Models
Let the model call external functions	tools parameter	Tool calling
Have the model output in a fixed JSON structure	response_format parameter	Structured outputs
Pass images to the model for understanding	multimodal content	Multimodal input
Stream responses token by token, lower time-to-first-token	stream: true	Streaming
AI music generation	Suno REST API	Suno API

Protocol compatibility scope

RouteAPI tries to preserve the native calling experience of each protocol while forwarding requests to the appropriate model service. Actual supported capabilities depend on the model, the service’s capabilities, and the request parameters:

Capability	Description
Chat Completions	The recommended basic chat endpoint, suitable for most OpenAI SDK compatible clients
Responses	Suitable for clients and coding agents that support the OpenAI Responses protocol
Embeddings	Used for vector search, semantic search, and RAG
Streaming	Returns incremental content using SSE
Claude Messages	Supports Claude’s native message structure, suitable for Claude Code and the Anthropic SDK
Google Gemini	Supports Gemini generateContent-style requests
Tool Calling	Depends on whether the model supports tool calling
Structured Outputs	Depends on whether the model supports JSON mode or JSON Schema
Vision / Multimodal	Depends on whether the model supports image or multimodal input

Capabilities are provided per model, not per platform: the same endpoint may lose tool calling or image input support if you switch model. Test with your target model before going live, or check the capability fields returned by Models.

API stability conventions

• Request parameters are preserved and forwarded according to the protocol whenever possible.
• If an optional scalar parameter is explicitly passed as 0 or false, RouteAPI treats it as an explicit value rather than dropping it as a default.
• Parameters unsupported by a given model may be adapted, ignored, or trigger an error, depending on that model’s compatibility rules.
• In production, we recommend pinning model IDs and preparing a fallback strategy for critical business flows.
• For optional capabilities like tool calling, structured outputs, visual input, and streaming usage statistics, validate them in a test environment before going live.

Enterprise integration recommendations

• Wrap your RouteAPI Token on the server side so business frontends never hold the key directly.
• Use different Tokens for different business systems, to make independent quotas, auditing, and troubleshooting easier.
• Pin model IDs and protocol paths — don’t rely on temporary aliases or display names.
• Log request ID, model ID, status code, latency, and token usage, so you can investigate latency and cost anomalies.
• For core business flows, enable streaming timeouts, failure retries, and fallback model plans to reduce the impact of a single model service disruption.

What’s next

• Authentication — create Tokens, configure quotas and permissions.
• Errors and debugging — status code meanings and a troubleshooting order.
• Migrating from OpenAI — a switchover checklist for existing applications.
• Client integration — specific configuration for Cursor, Claude Code, LangChain, and more.
• Billing and quotas — usage statistics and billing conventions.