diff --git a/main.py b/main.py
new file mode 100644
index 0000000..c2eef31
--- /dev/null
+++ b/main.py
@@ -0,0 +1,87 @@
+import random
+from typing import TypedDict, Dict
+from langgraph.graph import StateGraph
+from langchain_openai import ChatOpenAI
+from langchain_core.messages import HumanMessage, AIMessage
+
+# Define state
+class AgentState(TypedDict):
+    task: str
+    result: str
+    attempts: int
+    status: str  # pending | success | failed | max_attempts
+    error: str | None
+    max_attempts: int
+
+# Unreliable tool
+async def unreliable_tool(task: str) -> str:
+    if random.random() < 0.3:
+        raise ValueError("Tool failure")
+    # simple eval for arithmetic
+    try:
+        return str(eval(task))
+    except Exception as e:
+        raise ValueError(f"Eval error: {e}")
+
+# Execute task node
+async def execute_task(state: AgentState) -> Dict[str, str]:
+    try:
+        result = await unreliable_tool(state["task"])
+        return {"result": result, "error": None}
+    except Exception as e:
+        return {"result": "", "error": str(e)}
+
+# Verify result node using LLM-as-judge
+llm = ChatOpenAI(temperature=0)
+async def verify_result(state: AgentState) -> Dict[str, str]:
+    prompt = f"\nTask: {state['task']}\nResult: {state['result']}\nIs this correct? Answer with 'success' or 'failed'."
+    msg = await llm.ainvoke([HumanMessage(content=prompt)])
+    verdict = msg.content.strip().lower()
+    if "success" in verdict:
+        return {"status": "success"}
+    else:
+        return {"status": "failed"}
+
+# Handle error / retry node
+async def handle_error(state: AgentState) -> Dict[str, str]:
+    attempts = state["attempts"] + 1
+    if attempts >= state["max_attempts"]:
+        return {"status": "max_attempts", "attempts": attempts}
+    else:
+        return {"attempts": attempts, "status": "pending"}
+
+# Build graph
+builder = StateGraph(AgentState)
+builder.add_node("execute_task", execute_task)
+builder.add_node("verify_result", verify_result)
+builder.add_node("handle_error", handle_error)
+
+builder.set_entry_point("execute_task")
+builder.add_conditional_edges(
+    "execute_task",
+    lambda x: "verify_result" if x["error"] is None else "handle_error",
+)
+builder.add_conditional_edges(
+    "verify_result",
+    lambda x: "end_success" if x.get("status") == "success" else (
+        "end_max_attempts" if x.get("status") == "max_attempts" else "handle_error"
+    ),
+)
+builder.add_edge("handle_error", "execute_task")
+
+builder.set_finish_nodes(["end_success", "end_max_attempts"])
+
+graph = builder.compile()
+
+# Run demo
+if __name__ == "__main__":
+    initial_state: AgentState = {
+        "task": "2+2",
+        "result": "",
+        "attempts": 0,
+        "status": "pending",
+        "error": None,
+        "max_attempts": 5,
+    }
+    result = graph.invoke(initial_state)
+    print("Final state:", result)