mcp

Mozilla released the MDN MCP server, enabling AI agents and IDEs to access up-to-date MDN documentation and browser compatibility data via the Model Context Protocol, reducing reliance on outdated training data.

Explores the trend of AI agents autonomously spending money to complete real-world tasks like purchasing services, booking resources, and running ads without human approval.

WeChat released the developer access guide for Mini Program AI development mode (beta), allowing developers to encapsulate Mini Program functions as SKILLs via MCP protocol for AI invocation. It is currently in internal testing.

Centralaizer is an open-source, local memory hub that lets AI tools like Claude Desktop and Cursor share a persistent memory, with opt-in saving, PII scrubbing, and vector/full-text/knowledge-graph search.

Se lanzó Maple, una plataforma open source donde un agente de IA se conecta mediante MCP para encontrar errores y proponer soluciones en el backend.

OpenCut is an open-source video editor based on a Rust core, supporting cross-platform across Web/Desktop/Mobile. It provides innovative features such as plugin architecture, MCP server supporting AI agent calls, Headless mode, etc., aiming to replace paid video editing software.

Iris is an MCP server that runs inside your real app to verify AI agent work (e.g., Claude Code, Codex, Hermes) by checking conditions and returning a pass/fail verdict with evidence, reducing false positives and token usage compared to snapshot-based approaches.

A new dating app called 'Human in the Love' that operates as an MCP (Model Context Protocol) application inside Claude, claimed to be the first of its kind.

This article discusses the concept of Skills in the AI agent ecosystem, arguing that Skills are more than prompts—they are packaged capabilities that externalize human expertise into reusable workflow units. The author shares design principles and case studies from building popular Skills.

AI DevKit is a tool that adds a structured engineering workflow to AI coding agents, incorporating requirements, design, verification, and memory to replace one-off chats with repeatable processes.

Graperoot is an MCP-native tool that builds a dependency graph of a codebase to avoid unnecessary file re-reading, saving users significant costs—over $150k collectively—and is free for any CLI or IDE supporting MCP.

Nessie is a tool that allows users to transfer their context, memory, and history from ChatGPT, Perplexity, and Gemini to other platforms, including OpenClaw and the Hermes Agent, using OpenClaw and MCP servers.

Introduces an open-source learning project that implements a Claude Code-like agent harness from scratch, covering 20 lessons that gradually add core mechanisms, along with runnable code and documentation.

Discusses how to establish shared memory among multiple AI agents to avoid repeating mistakes, and introduces a solution by modifying the MemOS CLI to only record key information and search when necessary.

A new open-source AI tool implements Andrej Karpathy's LLM Council concept with Docker, MCP support, local/cloud models, and search integration for better multi-model deliberation.

The author built a code context graph parser that creates a graph from static analysis and exposes it via MCP for AI agents. In a head-to-head comparison with Gemma 4 26B, agents using the graph explored Apache Kafka's request flow in under 2 minutes, while the baseline agent without the graph ran out of rate limits in 6 minutes.

Fivos Aresti announces that his company has moved its entire operational knowledge to GitHub, connecting 25+ tools via MCPs to spin up contextual AI assistants instantly. The system features a markdown-based company OS, automated context propagation, self-improvement engines, and strict operating principles.

This paper introduces VATS, a mutation-driven framework that systematically evolves adversarial payloads to exploit error-path injection in MCP-based tool-calling agents. It demonstrates that error messages with implicit authority triple the success rate of standard indirect prompt injection across frontier models.

Aquifer is an MCP runtime that provides bounded queues, fairness controls, and dynamic pacing to handle rate limits and traffic spikes in AI agent systems. It also introduces the Aqueduct Protocol for dynamic flow state communication.

WeChat Agent has been officially released, capable of directly operating mini-program pages and combining with merchant-customized Skills and MCP services, marking the extension of WeChat's AI landscape into the real economy.