LangGraph + Connect AI
Build stateful graph-based agent pipelines with Connect AI as a queryable data node. Multi-node pipelines that discover data, analyze results, and generate outputs autonomously.
from langgraph.graph import StateGraph
from langgraph.prebuilt import create_react_agent
# 3-node pipeline: gather → analyze → render
gatherer = create_react_agent(llm, tools=mcp_tools)
graph = StateGraph(AgentState)
graph.add_node("gather", gatherer)
graph.add_node("analyze", analyze_node)
graph.add_node("render", render_node)
graph.set_entry_point("gather")
graph.add_edge("gather", "analyze")
graph.add_edge("analyze", "render")
app = graph.compile()
result = await app.ainvoke({"account": "Acme"})
Developer guide
Full step-by-step guide to building a stateful LangGraph agent pipeline with Connect AI MCP.
Sources for LangGraph
Common LangGraph + source patterns. Tap any combination to see the tutorial.
Sample agent
A polished LangGraph reference using the StateGraph pattern. Want a one-command start with LangChain, Google ADK, or Vercel AI SDK instead? See the npx scaffold.
3-node LangGraph pipeline (gather → analyze → render) that produces HTML customer health briefs from live data.
Evaluating Connect AI for your team? Learn more about Connect AI.
Full API reference, authentication guides, and configuration details.