LANGCHAIN

LangChain on 35+ models, one base_url

Pass Luno's base_url to ChatOpenAI and your chains, agents and RAG pipelines run on Claude, GPT and Gemini — one key, up to 70% cheaper in production.

GET YOUR API KEY Read the docs

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    base_url="https://api.luno.codes/v1",
    api_key="luno-...",
    model="claude-opus-4-8",
)

35+ models · OpenAI-compatible · up to 70% cheaper

// WHY LUNO HERE

One constructor argument

ChatOpenAI(base_url=..., api_key=..., model=...) is the entire integration — Python or JS, chains, agents, LCEL and streaming all work unchanged.

Swap models without rewrites

Change the model string from claude-opus-4-8 to gpt-5.5 or gemini-3.1-pro and rerun. No provider-specific classes, no second API account.

Production volume, flat cost

RAG pipelines and multi-agent graphs make thousands of calls per day. Luno's flat plans cut that spend by up to 70% versus metered per-token billing.

// QUICK SETUP

Get started in 30 seconds

Get your luno- key from the Telegram bot

Pass base_url and api_key to ChatOpenAI

Set any model ID — Claude, GPT or Gemini — and run your chain

// FAQ

FREQUENTLY ASKED QUESTIONS

Pass Luno's base_url to ChatOpenAI and your chains, agents and RAG pipelines run on Claude, GPT and Gemini — one key, up to 70% cheaper in production.

Yes. Both langchain_openai.ChatOpenAI (Python) and @langchain/openai (JS) accept base_url / configuration.baseURL — point them at https://api.luno.codes/v1 with your luno- key.

Yes — Luno also exposes an Anthropic-compatible endpoint, so ChatAnthropic works by overriding its base URL. Most users stick with ChatOpenAI since one class then covers every model.

Yes. bind_tools, with_structured_output, streaming and async all use standard OpenAI semantics, which Luno supports across Claude, GPT and Gemini models.

USE LUNO EVERYWHERE