All published AI Radar articles.

PentesterFlow/agent is an open-source, agentic AI command-line interface (CLI) designed for offensive security tasks. It assists penetration testers and bug hunters with recon, enumeration, validation, evidence collection, and reporting, while maintaining human oversight.

AgentCore is a Go library designed for building AI agent applications, emphasizing a minimal and composable core with extensibility. It supports single and multi-agent architectures, offering features like tool call gating, sub-agent invocation, context management with auto-summarization, and a unified event stream for lifecycle events.

Anthropic has published an analysis detailing insights gained from mapping AI-enabled cyber threats over the past year, focusing on their implications for cybersecurity.

MiniCode is a lightweight terminal-based coding assistant designed for local development. It features a Claude Code-like workflow, tool looping, and a Terminal User Interface (TUI) architecture. The project aims to be compact and easy to understand, making it suitable for learning and experimentation, with implementations available in TypeScript, Python, and Rust.

Aegis is an open-source Endpoint Detection and Response (EDR) tool designed to monitor the behavior of autonomous AI agents at the operating system level. It tracks processes, file system access, network activity, and applies behavioral analysis to detect anomalies without requiring agent-side hooks.

Researchers have introduced NewtPhys, a 4D physically annotated dataset designed to evaluate foundation models' understanding of low-level Newtonian physics. The dataset, built from multiview images of real-world scenes with physics-grounded simulations, provides detailed annotations including 3D forces and per-pixel quantities. Initial evaluations using NewtPhys on 56 Vision-Language Models (VLMs) and 10 Vision Foundation Models (VFMs) revealed limitations in their physics reasoning capabilities.

Researchers have identified two straightforward strategies to improve lidar semantic scene completion (SSC) performance without complex architectural changes: incorporating semantic pseudo-labels from existing segmentors and adding visibility information to lidar scans. These methods allow older models to achieve competitive or even superior results compared to current state-of-the-art systems.

AgentCrew is a Python-based chat application that enables the creation and management of multi-agent AI systems. It supports various AI models and the Model Context Protocol (MCP), allowing users to deploy specialized AI agents for tasks like coding, research, and automation.

Volcengine has released OpenViking, an open-source context database designed for AI agents. It unifies the management of agent context (memory, resources, and skills) using a file system paradigm, aiming to simplify context management and enable hierarchical context delivery and self-evolving capabilities.

A GitHub repository provides materials for a hands-on workshop focused on building multi-agent AI systems. It includes code, slides, and video for creating a Deep Research Agent and a Writing Workflow, both designed as MCP (Multi-Agent Communication Protocol) servers.