 |
VOOZH | about |
|
VOOZH | about |
LangGraph is a framework for building and visualizing intelligent, graph-based AI workflows that combine reasoning models (LLMs) with tool integrations and data operations. By integrating it with the CData Connect AI, you can enable agents to securely access, query, and act on live enterprise data in real-time via a standardized toolset.
CData Connect AI is a managed MCP-platform that allows you to expose your data sources (such as SAP SuccessFactors) through the Model Context Protocol (MCP). This means your AI agents can work with metadata, catalogs, tables, and SQL-enabled data access from hundreds of data sources, without complex ETL or custom integrations.
This article explores how to register the MCP endpoint in LangGraph, configure data source connectivity via CData Connect AI, and build a workflow that queries and visualizes live data (for example, SAP SuccessFactors objects) on demand. It demonstrates how to use the built-in MCP toolset (getCatalogs, getSchemas, getTables, queryData, etc.) to enable natural-language agents to interact with your enterprise data securely and interactively.
Before LangGraph can access SAP SuccessFactors, a SAP SuccessFactors connection must be created in CData Connect AI. This connection is then exposed to LangGraph through the remote MCP server.
You can authenticate to SAP Success Factors using Basic authentication or OAuth with SAML assertion.
You must provide values for the following properties to successfully authenticate to SAP Success Factors. Note that the provider will reuse the session opened by SAP Success Factors using cookies. Which means that your credentials will be used only on the first request to open the session. After that, cookies returned from SAP Success Factors will be used for authentication.
You must provide values for the following properties, which will be used to get the access token.
LangGraph authenticates to Connect AI using an account email and a Personal Access Token (PAT). Creating separate PATs for each integration is recommended to maintain granular access control.
With the SAP SuccessFactors connection configured and a PAT generated, LangGraph is prepared to connect to SAP SuccessFactors data through the CData MCP server.
Set up your project directory and install the required dependencies to connect LangGraph with CData Connect AI and use OpenAI LLM for reasoning. This setup enables LangGraph to call the SAP SuccessFactors MCP server tools exposed by Connect AI while OpenAI processes the natural language reasoning.
mkdir LangGraph cd LangGraph
pip install langgraph langchain-openai langchain-mcp-adapters python-dotenv "langgraph-cli[inmem]"
LangGraph uses environment variables to connect to the CData Connect AI and define the API credentials and configuration settings. Store these credentials in a .env file to keep them secure and reusable. LangGraph automatically reads this file at runtime, so the script can authenticate and communicate with the MCP server without hardcoding sensitive information.
# LangSmith (Optional) LANGSMITH_API_KEY=lsv2_pt_xxxx #LangSmith API Key LANGCHAIN_TRACING_V2=true LANGCHAIN_PROJECT=LangGraph-Demo # MCP Configuration MCP_BASE_URL=https://mcp.cloud.cdata.com/mcp #MCP Server URL MCP_AUTH=base64encoded(EMAIL:PAT) #Base64 encoded Connect AI Email:PAT OPENAI_API_KEY=sk-proj-xxxx
Note: You can generate the Base64-encoded authorization string using any online Base64 encoder, such as Base64 encoding tool. Encode your CData Connect AI username and PAT (obtained in the prerequisites).
In this step, you need to create a Python script that connects LangGraph to your CData Connect AI MCP server. The script retrieves the available MCP tools, such as getCatalogs, getSchemas, and queryData, builds a LangGraph workflow, and runs a natural language prompt against your connected SAP SuccessFactors data.
The workflow uses the MCP to securely retrieve live SAP SuccessFactors data from Connect AI and uses OpenAI GPT-4o to interpret and reason over that data. You also expose the graph so you can visualize it later in LangGraph Studio.
Create a new Python file named test.py inside your LangGraph project folder.
Use the following script into test.py:
import asyncio
import os
import operator
from typing_extensions import TypedDict, Annotated
from dotenv import load_dotenv
from langgraph.graph import StateGraph, START, END
from langchain.agents import create_agent
from langchain_mcp_adapters.client import MultiServerMCPClient
from langchain_openai import ChatOpenAI
from langchain_core.messages import BaseMessage, HumanMessage
# Load environment variables
load_dotenv()
# Define the agent state
class AgentState(TypedDict):
messages: Annotated[list[BaseMessage], operator.add]
# Define the async agent logic
async def run_agent(user_prompt: str) -> str:
MCP_BASE_URL = os.getenv("MCP_BASE_URL")
MCP_AUTH = os.getenv("MCP_AUTH")
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
print("Connecting to the MCP server...")
mcp_client = MultiServerMCPClient(
connections={
"default": {
"transport": "streamable_http",
"url": MCP_BASE_URL,
"headers": {"Authorization": f"Basic {MCP_AUTH}"} if MCP_AUTH else {},
}
}
)
print("Loading available MCP tools...")
all_mcp_tools = await mcp_client.get_tools()
print(f"Loaded tools: {[tool.name for tool in all_mcp_tools]}")
# Initialize the LLM
llm = ChatOpenAI(model="gpt-4o", temperature=0.2, api_key=OPENAI_API_KEY)
print("Creating the LangGraph agent...")
agent = create_agent(
model=llm,
tools=all_mcp_tools,
system_prompt="You are a helpful assistant. Use tools when needed."
)
# Build the workflow graph
builder = StateGraph(AgentState)
builder.add_node("agent", agent)
builder.add_edge(START, "agent")
builder.add_edge("agent", END)
graph_instance = builder.compile()
print(f"Processing user query: {user_prompt}\n")
initial_state = {"messages": [HumanMessage(content=user_prompt)]}
result = await graph_instance.ainvoke(initial_state)
print(f"Agent Response:\n{result['messages'][-1].content}")
# Expose the graph for visualization
builder = StateGraph(AgentState)
builder.add_node(
"agent",
create_agent(
model=ChatOpenAI(model="gpt-4o", temperature=0.2, api_key=os.getenv("OPENAI_API_KEY")),
tools=[],
system_prompt="You are a helpful assistant."
)
)
builder.add_edge(START, "agent")
builder.add_edge("agent", END)
graph = builder.compile()
if __name__ == "__main__":
import argparse
parser = argparse.ArgumentParser()
parser.add_argument("--serve", action="store_true", help="Run visualization server")
args = parser.parse_args()
if args.serve:
print("To visualize the graph, run:")
print("langgraph dev")
else:
asyncio.run(run_agent("List the first 2 catalogs available"))
Configure the LangGraph project so the CLI recognizes the workflow graph and environment settings. Create a configuration file that registers the graph for use in LangGraph Studio or during local visualization runs.
Create a new file named langgraph.json in your project directory.
Use the content below in the langgraph.json file:
{
"dependencies": ["."],
"graphs": {
"agent": "./test.py:graph"
},
"env": ".env"
}
Run the LangGraph development server to view and interact with your workflow in LangGraph Studio. This allows direct visualization of how the agent processes prompts, invokes tools, and retrieves SAP SuccessFactors data through the MCP server.
Open a terminal in your project directory and run:
langgraph dev👁 Start the LangGraph development server
After the server starts, LangGraph launches a local API and provides a link to the Studio UI:
https://smith.langchain.com/studio/?baseUrl=http://127.0.0.1:2024
Ideally, the link opens automatically when the command is run. If not, open this link in your browser to load the LangGraph Studio dashboard.
In the Studio interface, enter a natural language prompt such as:
Show all SAP SuccessFactors tables available in my catalog
LangGraph displays a real-time visualization of the agent's reasoning flow, showing how it interprets the prompt, calls the appropriate MCP tools, and retrieves live data from SAP SuccessFactors.
👁 Query and visualize the SAP SuccessFactors dataTo get live data access to hundreds of SaaS, Big Data, and NoSQL sources directly from your cloud applications, try CData Connect AI today!
Learn more about CData Connect AI or sign up for free trial access:
Free Trial
