360Labs.dev()
0%
Back to Portfolio
AI Builds|Live

AgentGuard

Deadlock prevention for multi-AI-agent systems -extends Dijkstra’s Banker’s Algorithm with progress monitoring, delegation cycle detection, and adaptive demand estimation.

Overview

AgentGuard is a library that brings formal deadlock prevention to multi-AI-agent systems by extending Dijkstra’s Banker’s Algorithm with three novel subsystems: Progress Monitor for silent stall detection, Delegation Tracker for authority cycle prevention, and Adaptive Demand Estimator for runtime resource pattern learning.

Specs

CoreExtended Banker’s Algorithm for LLM agents
ImplementationC++17 with Python bindings via pybind11
Tests285 tests (189 C++, 96 Python)
IntegrationNative LangGraph/LangChain support

Features

  • Mathematically guaranteed deadlock prevention
  • Progress Monitor with configurable stall detection
  • Delegation Tracker with real-time cycle detection
  • Adaptive Demand Estimator with statistical learning
  • Native LangGraph and LangChain integration
  • Open-source under MIT license on PyPI

Tech Stack

C++17Pythonpybind11LangGraphLangChain

Interested in AgentGuard?

Let's discuss how this can work for you.

Get in Touch