LangGraph Pre-Flight Check

LangGraph Agent Readiness

The pre-flight check for LangGraph applications. Detects StateGraph cycles, missing END conditions, and token consumption patterns.

Scan Your LangGraph AppView on GitHub

Common LangGraph Logic Flaws

Patterns that static analysis tools like linters don't catch.

Graph Cycles

StateGraph edges that create cycles without conditional routing to END

GraphRecursionError

Exceeding recursion_limit due to missing termination conditions

State Accumulation

Unbounded message lists in state growing with each graph step

Detection Patterns

LangGraph-specific detection patterns with code examples.

Graph Cycle Without END

CRITICAL

StateGraph edges create a cycle with no path to the END node.

Vulnerable
python
graph = StateGraph(AgentState)
graph.add_node("agent", call_agent)
graph.add_node("tools", call_tools)

# Cycle: agent -> tools -> agent (no END!)
graph.add_edge("agent", "tools")
graph.add_edge("tools", "agent")

graph.set_entry_point("agent")
app = graph.compile()
# Raises GraphRecursionError after 25 iterations
Secure
python
graph = StateGraph(AgentState)
graph.add_node("agent", call_agent)
graph.add_node("tools", call_tools)

# Conditional edge routes to END when done
graph.add_conditional_edges("agent", should_continue, {
    "continue": "tools",
    "end": END
})
graph.add_edge("tools", "agent")

graph.set_entry_point("agent")
app = graph.compile()

Non-Deterministic Exit

HIGH

Loop termination depends on LLM output, which may never signal completion.

Vulnerable
python
def should_continue(state):
    # LLM decides - non-deterministic!
    last_message = state["messages"][-1]
    if "DONE" in last_message.content:
        return "end"
    return "continue"  # May never return "end"
Secure
python
def should_continue(state):
    last_message = state["messages"][-1]
    # Deterministic fallback after N steps
    if len(state["messages"]) > 20:
        return "end"
    if "DONE" in last_message.content:
        return "end"
    return "continue"

Getting Started

Run Inkog against your LangGraph codebase.

1

Run the scanner

bash
npx -y @inkog-io/cli scan ./my-langgraph-app
2

Review findings

Inkog traces data flow through your LangGraph code and reports issues with severity levels and line numbers.

3

Address issues

Apply the suggested fixes based on severity and re-scan to verify.

LangGraph Compliance Reports

Automated mapping to global AI governance frameworks.

EU AI Act

Article 14, 15, 12

NIST AI RMF

MAP/MEASURE/MANAGE

OWASP LLM

Top 10 Coverage

ISO 42001

AI Management

LangGraph Readiness FAQ

Does Inkog detect GraphRecursionError?

Yes. Inkog analyzes your StateGraph definition and identifies cycles that will trigger GraphRecursionError. It finds the issue at development time, not runtime.

How does Inkog analyze LangGraph?

Inkog has a dedicated LangGraph adapter that parses StateGraph definitions, add_node, add_edge, and add_conditional_edges calls. It builds the graph topology and detects cycles, missing END routes, and state accumulation.

Does Inkog support MessageGraph?

Yes. Both StateGraph and MessageGraph are supported. The adapter understands both patterns and their specific termination requirements.

Scan Your LangGraph Application

Free tier available. No credit card required.

Start Free ScanLearn More About Inkog