agentsentinel-langgraph

Security scanner for LangGraph agent workflows.
Powered by AgentSentinel on SingularityNET mainnet.

Install

pip install agentsentinel-langgraph

Quick Start

from langgraph.graph import StateGraph, START, END
from agentsentinel_langgraph import scan_graph, SecureGraph

builder = StateGraph(MyState)
builder.add_node("agent", my_agent_fn)
builder.add_node("tools", tool_node)
builder.add_edge(START, "agent")

# Scan builder before compiling
report = scan_graph(builder)
report.print()

# Or compile then scan + invoke securely
graph = builder.compile(checkpointer=MemorySaver())
secure = SecureGraph(graph, block_on="CRITICAL")
secure.scan()

# Use graph normally after scan
result = graph.invoke({"messages": [HumanMessage(content="hello")]})

What It Checks

• AGT04 — No human-in-the-loop interrupt points (fully autonomous)
• AGT08 — High recursion limit / infinite loop risk
• AGT03 — No checkpointer (cannot audit or rollback)
• AGT02 — Shell/exec nodes in graph
• AGT05 — Credential patterns in config

Links

• AgentSentinel on SingularityNET
• Contact: medhasni760@gmail.com

MIT License — © 2026 NeuralSentinel