Agent Protocols news and analysis on Web Pulse: 3169 curated articles tracking the latest Agent Protocols developments, tools, and research, updated continuously from vetted sources.
Chrome 155 and Chrome 156 add a visual position-area editor for CSS anchor positioning in DevTools and a new get_css_styles MCP tool that lets AI agents read page styles, byteiota reported. The get_css_styles tool extend…
Cole Medin, a developer who runs an AI "second brain" built on Claude Code with Obsidian as a Markdown memory vault, reports the lean setup saves him 25–30 hours weekly, according to a first-person account of the archite…
A developer built a free, no-account llms.txt generator that reads a site's sitemap and produces a spec-style /llms.txt file in seconds, plus an optional llms-full.txt containing short excerpts of every page. The tool fo…
A developer built Blink Artifacts, a remote MCP service that lets any MCP client publish a shareable web page which humans edit in the browser while an agent keeps editing it over MCP. Each artifact is a Cloudflare Durab…
A developer guide shows how to connect GitHub's official MCP Server to Copilot in VS Code, letting the agent read issues, pull request review comments, GitHub Actions failures, security findings and Dependabot alerts dir…
Upstash's Context7 pulls current, version-specific library documentation and code examples into coding agents like Cursor, Claude Code, VS Code and Codex via MCP, aiming to stop agents from writing against outdated or in…
Google released Apigee hybrid 1.17.0 on September 14, 2026, adding Model Context Protocol support so the MCP data plane runs inside a customer's own Kubernetes cluster. The release follows the December preview of MCP sup…
Sparetext launched an iPhone SMS gateway that sends texts from a user's own carrier number by relaying through a Mac running macOS Text Message Forwarding and AppleScript, and it ships a hosted MCP server so AI agents ca…
A step-by-step guide details how to connect Qwen Audio Agent, an open source voice front end, to Claude Code as the coding back end through a local gateway and an adapter for the agent client protocol (ACP), enabling spo…
The Model Context Protocol has finalized SEP-2133's successor, a SEP that gives MCP extensions a three-stage maturity lifecycle — Experimental, Beta and Stable — with the stage recorded in the repository path (`specifica…
MotherDuck launched Dive Apps, fully custom data apps that run on the MotherDuck data warehouse and can write back to external systems through Flights, MotherDuck's serverless Python runtime. Dive Apps are built by conne…
Google announced Gemini 4 Carbon on 28 Oct 2026, describing it as "our strongest model with state-of-the-art cybersecurity and long horizon capabilities," with introductory pricing of $2 per million input tokens and $10 …
The claude-mem persistent memory plugin for Claude Code is trending on GitHub this week with about 104,000 stars, according to its repository README, which lists version 15.5.0 under an Apache-2.0 license. The plugin hoo…
Independent researcher Syed Anas Mohiuddin found the same server-side request forgery (SSRF) flaw in MCP servers run by Google, JPMorgan Chase, Weaviate, France's interministerial digital directorate and the city governm…
Mintlify launched Mintlify Base in beta, an internal-knowledge workspace built on the same platform as Mintlify Docs where people and agents share specs, runbooks, and onboarding pages. Base includes a built-in MCP serve…
AI educator Nate Herk popularized a four-component framework for building a personal "AI second brain" — context, connections, cadence, and capabilities — typically run on a coding agent such as Claude Code plus a knowle…
TikTok's Privacy Innovation Lab published a paper on February 27, 2026, by Sirat Samyoun and Jian Du, titled "When Tools Become Prompts: Why Multi-Turn MCP Systems Break Traditional Security Assumptions," showing that a …
Anthropic's Applied AI team published guidance arguing that enterprise agent architectures should default to a single agent loop with skills and tools loaded on demand, reserving subagents only for cases needing isolated…
A developer published a 10-minute review checklist for approving MCP servers before granting them access to an AI assistant, warning that tool descriptions are read by the model as instructions and that servers inherit t…
Agnostic-ai released v0.90.0, adding per-tool settings for model and effort through the Settings > What each tool gets interface and tightening sync safety around project paths, pack updates, and imported credentials. Th…