langchain-leanroute

LangChain integration for Leanroute, the agent-native AI gateway.

ChatLeanroute is LangChain's ChatOpenAI pointed at Leanroute: one key for 100+ models across 14 providers (OpenAI, Anthropic, Google, xAI, DeepSeek, Qwen and more), with failover, caching, spend caps, MCP tools on any model, and cost per agent run. Tools, structured output, streaming, async and LangGraph work unchanged.

Install

pip install langchain-leanroute

Create a runtime key (gw_live_*) at leanroute.dev/dashboard/keys and set it:

export LEANROUTE_API_KEY=gw_live_...

New accounts get 100 free requests on low-cost models, no card needed.

Use

from langchain_leanroute import ChatLeanroute

llm = ChatLeanroute(model="deepseek/deepseek-v4-1-flash")
print(llm.invoke("Explain retrieval-augmented generation in one sentence.").content)

Model ids use the provider/model form, e.g. anthropic/claude-sonnet-5 or openai/gpt-5.6-terra. Full list: leanroute.dev/docs/models.

Agent runs

Name your agent and give each run an id, and Leanroute groups the calls into runs on the Agent runs page: cost per run, models used, and every tool called.

from langchain_core.tools import tool
from langgraph.prebuilt import create_react_agent
from langchain_leanroute import ChatLeanroute, new_run_id


@tool
def get_weather(city: str) -> str:
    """Return the current temperature for a city."""
    return f"{city}: 22°C"


def run(task: str) -> str:
    llm = ChatLeanroute(
        model="anthropic/claude-sonnet-5",
        agent_name="weather-agent",
        session_id=new_run_id(),  # one per run
    )
    agent = create_react_agent(llm, tools=[get_weather])
    result = agent.invoke({"messages": [("user", task)]})
    return result["messages"][-1].content


print(run("What's the weather in Singapore?"))

agent_name is sent as X-Agent-Name (without it, the API key's name is used) and session_id as X-Agent-Session-Id. Headers are fixed per instance, so create one ChatLeanroute per run. Every request also carries X-Leanroute-Client: langchain-leanroute/<version>.

Options

Every other ChatOpenAI argument (temperature, max_tokens, timeout, max_retries, ...) passes through.

Links

• LangChain + Leanroute guide
• Agent runs
• Gateway headers

License

Apache-2.0. See LICENSE.