For AI Engineers

Debug the Logic.

Ship Agents that Sleep at Night.

The Trace

Visualize agent execution paths. Find bugs before your users do.

Path analysis
Call graph visualization
Cycle detection

Execution Path Visualization

Flame graph showing infinite recursion detection

main()
run_agent()
process_query()
call_llm()
handle_response()
needs_clarification()
recursive_loop()
recursive_loop()
recursive_loop()
Normal
Infinite Recursion

LLM Eval vs Static Analysis

Why deterministic testing wins

AspectLLM EvalInkog Static
Speed
2-10 seconds
<100ms
Results
Flaky, non-deterministic
Deterministic
Coverage
Sample-based
Complete path analysis
Cost
API calls every run
Zero runtime cost

Fix GraphRecursionError

Detect circular dependencies and infinite recursion in LangGraph workflows before runtime.

Trace Logic Flow

Understand complex agent decision trees with static flow analysis that maps every possible execution path.

Multi-Agent Security

Audit CrewAI, Swarm, and LangGraph for infinite delegation loops and privilege escalation between agents.

The Fix

agent.py
while agent.needs_clarification():
    response = agent.ask_user()
    agent.process(response)
    # No exit condition!
    # Can loop forever
Infinite loop detected - Blocked
Terminal
$ inkog scan agent.py
✓ Scanning agent code...
✗ ISSUE: Potential infinite loop between `ask_user()` and `process()`
→ Add exit condition or iteration limit
✓ 1 issue found

Why AI Engineers Choose Inkog

Instant Bug Detection

Find infinite loops and recursion errors without running your agent.

Token Cost Protection

Catch unbounded API calls before they rack up bills.

Framework Agnostic

LangChain, LangGraph, AutoGen, CrewAI – works with any stack.

Frequently Asked Questions

What is GraphRecursionError?+

Occurs when a LangGraph agent enters an infinite loop due to circular dependencies or missing exit conditions. Crashes your agent and drains API budget. Inkog catches these statically before runtime.

How does Inkog detect infinite loops?+

Builds a Data Flow Graph of your code and analyzes all execution paths. Identifies cycles without termination, recursive calls that overflow, and loops that never exit—without running your code.

What frameworks does Inkog support?+

LangGraph, LangChain, AutoGen, CrewAI, DSPy, Haystack. Framework-specific rules catch common mistakes like missing exit conditions or unbounded tool loops.

How do I audit multi-agent delegation chains?+

Use the MCP server integration. Ask your AI assistant: 'Audit my CrewAI agents for security issues'. Inkog extracts all agents, maps delegation relationships, and detects infinite loops like Agent A → B → A.

Ready to ship reliable agents?

Get Started with Docs