Recipe: Basic LangGraph + RunProof

pip install substr8-langgraph langgraph

from langgraph.graph import StateGraph, END
from substr8_langgraph import instrument_graph, RunProofConfig
 
# Define your state
class AgentState(dict):
    messages: list
 
# Define nodes
def research_node(state):
    # Your research logic here
    return {"messages": state["messages"] + ["Researched topic"]}
 
def summarize_node(state):
    # Your summarization logic
    return {"messages": state["messages"] + ["Summary: ..."]}
 
# Build the graph
workflow = StateGraph(AgentState)
workflow.add_node("research", research_node)
workflow.add_node("summarize", summarize_node)
workflow.add_edge("research", "summarize")
workflow.add_edge("summarize", END)
workflow.set_entry_point("research")
 
# Compile
graph = workflow.compile()
 
# Instrument with RunProof
instrumented = instrument_graph(
    graph,
    config=RunProofConfig(
        agent_id="cookbook/research-agent",
        output_path="./runproof.json"
    )
)
 
# Run it
result = instrumented.invoke({"messages": ["Research AI governance"]})
 
print("✅ RunProof generated: ./runproof.json")

substr8 proof verify runproof.json

RunProof Verification
=====================
Proof ID: proof_abc123
Agent: cookbook/research-agent
Runtime: langgraph

Checks:
  ✓ Linear integrity (hash chain valid)
  ✓ Set integrity (merkle root matches)
  ✓ Signature valid

Status: VALID ✅