ToRouter DocsIntegrations
LangChain & LlamaIndex
Use LangChain's ChatOpenAI and LlamaIndex's OpenAI wrapper against ToRouter to mix Claude, GPT and Gemini in one chain.
LangChain's ChatOpenAI and LlamaIndex's OpenAI LLM wrapper both speak the OpenAI Chat Completions protocol — point them at ToRouter and you can drop any model from the catalog into chains, agents and RAG pipelines without changing framework code.
Configuration
| Field | Value |
|---|---|
| Base URL | https://portal.torouter.ai/v1 |
| API key | sk-*** |
| Model | claude-opus-4-7, gpt-5, gemini-2.5-pro, ... |
OPENAI_BASE_URL and OPENAI_API_KEY env vars are auto-detected by both libraries.
LangChain
from langchain_openai import ChatOpenAI
llm = ChatOpenAI(
base_url="https://portal.torouter.ai/v1",
api_key="sk-***",
model="claude-opus-4-7",
)
print(llm.invoke("Hello").content)import { ChatOpenAI } from '@langchain/openai';
const llm = new ChatOpenAI({
configuration: { baseURL: 'https://portal.torouter.ai/v1' },
apiKey: 'sk-***',
model: 'claude-opus-4-7',
});
const res = await llm.invoke('Hello');
console.log(res.content);For embeddings use OpenAIEmbeddings with the same base URL and an embedding model id from the catalog (e.g. text-embedding-3-large).
LlamaIndex
from llama_index.llms.openai import OpenAI
llm = OpenAI(
api_base="https://portal.torouter.ai/v1",
api_key="sk-***",
model="gpt-5",
)
print(llm.complete("Hello").text)Gotchas
- LangChain Python uses
base_url; LangChain JS usesconfiguration.baseURL; LlamaIndex usesapi_base. The values are identical, the field names are not. - When mixing models in one chain (e.g. Claude for reasoning, GPT for tool-use), instantiate one wrapper per model — don't try to swap
modelmid-call. - Some agent frameworks expect strict OpenAI tool-call schemas; if a non-OpenAI model produces tool calls in a slightly different format, prefer the Anthropic SDK directly for that step. See anthropic.