LangChain + CaMeL Hub: One Key for GPT, Claude, Gemini & DeepSeek

LangChain is the most widely used Python/JS framework for building LLM applications — chains, agents, RAG pipelines — on top of a swappable chat-model interface. Point its OpenAI-compatible ChatOpenAI class at CaMeL Hub and every chain, agent, or tool you already wrote keeps working unchanged, except now model="..." can be GPT-5.5, Claude Sonnet 5, Gemini 3.1 Pro, DeepSeek V4 Flash, or any of CaMeL Hub's 480+ models — behind one API key, one base URL, and one bill.

LangChain ↗

1. Install the LangChain OpenAI integration

   LangChain talks to any OpenAI-compatible endpoint — including CaMeL Hub — through the langchain-openai package. Install it alongside the base langchain package.

   pip install -U langchain langchain-openai

2. Create a CaMeL Hub API key

   Sign in to the CaMeL Hub console and open the page usually labeled Tokens or API Keys in the left sidebar. Create a new token and copy the sk-... string immediately — it is shown in full only once.

3. Point the OpenAI client config at CaMeL Hub

   LangChain's OpenAI classes read two settings: the API key and the base URL. Set them as environment variables (or pass them directly to ChatOpenAI in the next step). CaMeL Hub's OpenAI-compatible endpoint needs the full path with /v1 — the bare host https://api.camel-hub.com will not work here.

   export OPENAI_API_KEY="sk-your-camel-hub-key"
   export OPENAI_BASE_URL="https://api.camel-hub.com/v1"

4. Call any model with ChatOpenAI

   Import ChatOpenAI from langchain_openai and set model to any CaMeL Hub model slug. No extra base_url argument is required if you already set the environment variables above, but you can also pass it explicitly.

   from langchain_openai import ChatOpenAI
   
   llm = ChatOpenAI(
       model="claude-sonnet-5",
       base_url="https://api.camel-hub.com/v1",  # optional if OPENAI_BASE_URL is set
       temperature=0.7,
   )
   
   response = llm.invoke("Explain what a base_url override does, in one sentence.")
   print(response.content)

5. Switch providers by changing only the model string

   Because every provider is exposed through the same OpenAI-format endpoint, you never need LangChain's ChatAnthropic or ChatGoogleGenerativeAI classes here — just change the model string and keep using ChatOpenAI.

   for model in ["gpt-5.5", "claude-sonnet-5", "gemini-3.1-pro", "deepseek-v4-flash"]:
       llm = ChatOpenAI(model=model)
       print(model, "->", llm.invoke("Say hello in five words.").content)

6. Stream tokens and use tool calling

   Streaming and tool/function calling work exactly like they do against OpenAI directly. Use .stream() for token-by-token output, and .bind_tools() for agents — check the Capabilities section on a model's CaMeL Hub page first to confirm it supports function calling or vision.

   for chunk in llm.stream("Write a two-line haiku about API gateways."):
       print(chunk.content, end="", flush=True)

7. Same pattern in LangChain.js (TypeScript)

   The Node/TypeScript SDK follows the same pattern through @langchain/openai, so LangGraph.js agents and chains built on it need no other changes.

   import { ChatOpenAI } from "@langchain/openai";
   
   const llm = new ChatOpenAI({
     model: "claude-sonnet-5",
     apiKey: process.env.OPENAI_API_KEY,
     configuration: { baseURL: "https://api.camel-hub.com/v1" },
   });
   
   const res = await llm.invoke("hello");
   console.log(res.content);

Frequently asked questions