All published AI Radar articles.

The Agent Runtime Control Protocol (ARCP) has released its F# / .NET reference SDK, providing tools for submitting, observing, and controlling long-running AI agent jobs. The SDK, versioned v1.0.1, supports both client-side job submission and runtime-side agent hosting.

MateClaw, an open-source AI agent platform built on Spring AI Alibaba, has released version 1.3.0. This update introduces multi-agent orchestration via a new Workflow engine, event-driven Triggers, and an enhanced Wiki that acts as a processing pipeline for knowledge. It also improves tool binding for individual agents and adds multimodal sidecar routing for vision models.

Limpid is a new macOS-native terminal application designed for the era of AI coding agents. Built on the `libghostty` core, it offers a SwiftUI-based interface with features like project and worktree management, and dedicated support for parallel AI agent sessions.

Uni-CLI is an open-source operations substrate designed for AI agents to interact with various software environments, including websites, logged-in browsers, desktop applications, and local tools. It provides a unified execution layer for agents, offering features like searchable commands, policy-gated execution, evidence recording, and self-repair mechanisms.

Gnoma is an agentic coding assistant written in Go that offers provider-agnostic routing across various cloud (Anthropic, OpenAI, Gemini, Mistral) and local (Ollama, llama.cpp) large language models. It features a multi-armed bandit router for dynamic model selection, extensibility via MCP, skills, hooks, and plugins, and includes privacy-focused features like secret scanning and a no-phone-home policy.

cfgaudit is an open-source security auditor designed to identify misconfigurations in AI agent configuration files, initially focusing on Claude Code and Cursor. It helps developers ensure their AI coding assistants adhere to the principle of least privilege and prevent sensitive data exposure.

SpecAnchor has released v0.5.0-beta.1, reframing itself as a "Harness Context Control plane" for AI coding agents. This update introduces persistent management for Decision and Evidence contexts, alongside existing Spec Context, to improve agent performance and continuity across sessions by organizing and filtering crucial information.

Adam is an open-source, 5-layer persistent memory and identity architecture for AI agents, designed to ensure an agent's continuity and identity sovereignty. It stores an AI's core memory and neural graph in user-owned files, making it resilient to model swaps, vendor shutdowns, and machine wipes. The framework has been production-validated over 353 sessions and claims to be the first documented case of emergent values in persistent AI, quantum-verified on IBM hardware.

Comis is an open-source, self-hosted platform that enables AI agent teams to operate within various messaging applications like Telegram, Discord, Slack, and WhatsApp. It emphasizes security through kernel-enforced sandboxes and supply chain integrity, alongside cost efficiency via prompt caching.

Li+ is a new high-level programming language designed for AI agents, focusing on "prompt-native harness engineering." It enables AI to function as an interactive compiler, autonomously managing development tasks like issues, specifications, implementation, CI, and releases through human dialogue. The language emphasizes observable behavior as the definition of correctness and distills requirements from conversation.