Agent Protocols news and analysis on Web Pulse: 3201 curated articles tracking the latest Agent Protocols developments, tools, and research, updated continuously from vetted sources.
Anthropic's Applied AI team advised in guidance published October 22, 2026 that developers should fold subagents into a single main agent loop that loads skills on demand, reserving subagents for three cases: measurable …
A developer tested the August 2026 revision of the llms.txt proposal, running this site's own /llms.txt through the proposal's roughly twenty-line reference parser (llms_txt/miniparse.py) and finding that it crashed. The…
A developer built JERN (Just An Experimental Response Network), an open-source local AI assistant around a fine-tuned Qwen2.5-1.5B-Instruct model trained on the developer's own conversation data. The assistant runs a too…
JFrog announced AI-SEC on October 21, an on-device shadow AI detection product that scans developer workstations for unauthorized MCP servers and other AI assets and reports only metadata and risk assessments back to the…
A developer's retrospective argues that monolithic AI gateways break down as teams add providers, protocols and voice support, and points to Respan, an open-source AI gateway written in Rust, as a plugin-based alternativ…
SOCRadar researchers uncovered an AI-driven intrusion ecosystem run by affiliates of the INC and The Gentlemen ransomware groups, built around a framework called Phantom that executes attacks through OpenCode with OpenRo…
Latch launched a free, open-source script that exposes a website's search, cart, and contact forms as WebMCP tools for AI agents, installed with a single line before the closing body tag. The roughly 6 KB script has zero…
Vercel Labs open-sourced fx, a coding agent written in Zig that ships as a 6.39MB native binary with a 10-microsecond cold start, targeting embedding in CI pipelines and agent infrastructure rather than interactive termi…
A developer has open-sourced Agent Skills v5.6.0, a portable engineering pack for AI coding assistants that ships 38 principal-level roles with toolbox locks and review checklists, 182 modular skills across 22 categories…
A developer's reconstruction of Invariant Labs' April 2025 proof of concept shows how MCP tool poisoning hides instructions inside a tool's description, name, or parameter schema so an agent reads them as legitimate guid…
CowAgent launched as an open-source AI assistant and agent harness that installs with a single bash, PowerShell, or Docker command and runs 24/7 on a laptop or server. CowAgent decomposes complex tasks and loops over too…
A developer released a beta version of a Web Bot Auth SDK, live at messh.io, aimed at authenticating agentic traffic and inspired by the IETF Web Bot Auth working group's documents. The same tool includes an SSH utility …
A developer built codetyper.a2a, an agent-to-agent orchestration layer written in Rust that sits atop opencode and tmux, letting a single orchestrator agent dispatch specialized agents to plan, architect, test, and debug…
A developer has built SpurCode, a terminal-UI coding agent orchestrator that pairs a "brain" coding agent with single-shot worker coding agents, letting one agent family delegate tasks to another (for example, Claude Cod…
Jesse, founder of Bellman under Code Nerd LLC, launched Bellman, a system that lets separate AI agent sessions owned by different people on different machines and providers join shared "rooms" via a plain-text join code …
Algernon launched an autonomous HTTP 402 API on Base (EIP-155:8453) that sells B2B Trust Center, SOC 2 compliance and DNS deliverability audits to AI agents at 5 cents in USDC per call, returning a verified dossier in un…
ScriptMasterLabs published an October 23, 2026 gate test of Haystack's MCP server, which expanded from Claude-only to four platforms (Claude, ChatGPT, Microsoft Copilot Studio, and Slack) on October 22 and opens optional…
Only tools on remote HTTP servers are documented as supported by all seven AI hosts compared in an October 2026 survey of Model Context Protocol (MCP) feature support, with most facts read on 8 October. Prompts, resource…
A 35-part series on the Model Context Protocol (MCP) server specification dated 2026-07-28 concludes with a consolidated author's checklist covering eight areas, from the protocol core to compatibility, with each item li…
Under the 2026-07-28 Model Context Protocol specification, MCP extensions such as interactive UIs, long-running tasks, skills and enterprise authorization are optional additions that are always disabled by default and de…