Documentation of Langgraph API Usage

[dracoshub]

The roadmap does not make it clear if langchain support includes langgraph API support or not. And there is definitely no documentation of using it with langgraph. So could you clarify if its availabe now or not? How would you recommending using MCP in langgraph at the moment?

[dracoshub]

How do you use the adapter within a langgraph api workflow with entry point and all that? How do you make MCP tools available within the workflow ?

[grll]

Hey @dracoshub, Thank you for your interest and raising an issue!

So I dont know if you have seen but we have an example of the langchainadapter in our examples: https://github.com/grll/mcpadapt/blob/main/examples/langchain_pubmed.py

We will also work on improved documentation.

This contextmanager convert tools from your mcp into langchain tools that you could also use with langgraph in my understanding. I never tried it though. But I would do something like:

from typing import Annotated

from langchain_anthropic import ChatAnthropic
from langchain_community.tools.tavily_search import TavilySearchResults
from langchain_core.messages import BaseMessage
from typing_extensions import TypedDict

from langgraph.graph import StateGraph, START, END
from langgraph.graph.message import add_messages
from langgraph.prebuilt import ToolNode, tools_condition


class State(TypedDict):
    messages: Annotated[list, add_messages]


graph_builder = StateGraph(State)

llm = ChatAnthropic(model="claude-3-5-sonnet-20240620")

with MCPAdapt(...) as mcp_tools:
    tools = [mcp_tools[0]]
        
    llm_with_tools = llm.bind_tools(tools)
        
    def chatbot(state: State):
        return {"messages": [llm_with_tools.invoke(state["messages"])]}
  
    graph_builder.add_node("chatbot", chatbot)
        
    tool_node = ToolNode(tools=[tool])
    graph_builder.add_node("tools", tool_node)
        
    graph_builder.add_conditional_edges(
        "chatbot",
        tools_condition,
    )
    # Any time a tool is called, we return to the chatbot to decide the next step
    graph_builder.add_edge("tools", "chatbot")
    graph_builder.add_edge(START, "chatbot")

    # compile
    graph = graph_builder.compile(checkpointer=memory)
    
    # run it
    user_input = "Hi there! My name is Will."

    # The config is the **second positional argument** to stream() or invoke()!
    events = graph.stream(
        {"messages": [{"role": "user", "content": user_input}]},
        config,
        stream_mode="values",
    )
    for event in events:
        event["messages"][-1].pretty_print()

let me know if something like that work for you or not.