mcp

Introduces Apple Health MCP Server, an open-source Model Context Protocol server that enables SQL querying and DuckDB-based analysis of exported Apple Health data, featuring natural-language queries, health reports, and lazy loading.

Kent C. Dodds announces Kody is moving to heykody.app, positioning it as an AI agent that integrates with existing assistants like Cursor and Claude Code to securely access services and run automated jobs such as Slack catch-up, week planning, and GitHub triage.

Agentic Harness is a new open-source Python infrastructure that turns YAML files into executable agent workflows, featuring markdown audit logs, pre-built specialists, cost guardrails, native MCP support, and local model integration.

The Agent Plugins standard, backed by OpenAI, Cursor, Google, Vercel, and others, is released as a way to package agent skills with MCP connectors for easy installation on any harness. The author notes it is mostly packaging rather than a breakthrough, but important for agent distribution.

Raises concerns that MCP servers may introduce new architectural dependencies, questioning whether agents tied to specific server auth and implementations are truly portable.

Google has joined Open Agent Plugins as a core maintainer, helping build an agent ecosystem layer that covers discovery, cataloging, packaging, and running of portable agent capabilities.

Vercel announces Agent Plugins 1.0.0, an open, vendor-neutral standard for packaging Agent Skills and MCP servers into distributable plugins, giving AI agents a common format for discovery and loading.

OpenAI announces Agent Plugins, an open standard developed with AWS, Cursor, GitHub, Code, and Vercel to package Agent Skills and support MCP server configurations, enabling plugins to work across compatible agent clients. Matt Pocock reacts enthusiastically and expects a skills plugin soon.

Vercel and partners introduce Agent Plugins, an open standard for extending AI agents with skills and MCP servers, now supported by Cursor.

Kent C. Dodds has updated Kody Koala to the latest MCP spec, praising Cloudflare for easing the process and highlighting the rewritten stateless core that runs on Workers.

Kody v2026.08.06 is out, introducing new Standard ($5/mo) and Pro ($20/mo) pricing tiers, with the old Pro tier now becoming Standard. Kody is an MCP-based personal assistant platform built on Cloudflare Workers.

Blueprint MCP is an open-source MCP server that uses Nano Banana Pro to generate architecture diagrams, sequence diagrams, and flowcharts from codebases and system data, helping developers visualize and document complex systems.

Dembrandt is an open-source CLI tool that extracts a website's design system—logos, colors, typography, borders, spacing, and more—into design tokens. It supports MCP integration for AI agents and offers a companion app for drift tracking and visual diffs.

The author built an MCP tool that lets AI agents retrieve a Google Maps reviewer's full history and metadata for vetting reviews and detecting fraudulent patterns.

Wispr Flow introduces Wispr Notetaker, a meeting note tool that integrates with AI assistants like Claude, ChatGPT, and Cursor via MCP, turning transcripts into queryable context.

Wispr Flow's Notetaker integrates with AI agents via MCP to remember and retrieve exact conversations from past meetings, eliminating the need to dig through transcripts.

An opinion piece arguing that MCP servers often fail because teams build them without confirming real user demand, and recommends validating with actual struggling users and early usage instrumentation before committing.

Cloudflare is open sourcing Cloudflare OS, an open platform that gives every employee an AI agent and workspace grounded in company context, skills, and internal systems.

PullMD is an open-source, self-hostable service that converts web pages, PDFs, Office documents, EPUBs, images, audio, and YouTube videos into clean Markdown, with a REST API, MCP server, and PWA frontend.

This paper investigates when hybrid computer-use agents actually choose to use MCP tools versus screenshots, finding that tool availability alone does not guarantee adoption: a reasoning model improves while a non-reasoning model degrades. It also explores training and context compression strategies to close the adoption gap and reduce token costs.