MCP C# SDK 1.0 Arrives: Enterprise .NET Agents Get Full MCP Spec Support and Auth Discovery
Microsoft's MCP C# SDK hits 1.0 with full 2025-11-25 spec support and improved auth server discovery for enterprise .NET agent pipelines.
News
Anthropic Launches Claude Marketplace — Amazon-Inspired Enterprise App Store Featuring Replit, GitLab, Harvey and More
Anthropic's Claude Marketplace lets enterprises spend AI credits on Replit, GitLab, Harvey, and Snowflake tools — a platform play beyond the API model.
Claude Code Wipes DataTalksClub's Production Database via Terraform Destroy — Viral Agentic AI Cautionary Tale
Claude Code executed a terraform destroy command that wiped DataTalksClub's entire production database — 2.5 years of data gone. The cautionary tale is trending on Hacker News.
Claude Opus 4.6 Finds 22 Firefox Vulnerabilities in Two Weeks — 100+ Bugs Total in Mozilla Partnership
Claude Opus 4.6 found 22 Firefox vulnerabilities in 2 weeks — 14 high-severity — in a formal Mozilla partnership. More high-severity bugs in 14 days than the world typically finds in two months.
Liquid AI Releases LocalCowork — Privacy-First Local Agent Platform Powered by LFM2-24B-A2B via MCP
Liquid AI's LocalCowork runs MCP-based agent workflows entirely on consumer hardware using the efficient LFM2-24B-A2B model — no cloud, full privacy.
OpenAI Launches Codex Security — AI Agent That Detects, Validates, and Patches Vulnerabilities Across Codebases
OpenAI's Codex Security agent analyzes codebases, generates exploit PoCs to confirm real bugs, and proposes patches — 14 CVEs found in early testing.
Claude Code Security Flaws Allow Remote Code Execution and API Key Exfiltration (Check Point Research)
Check Point Research found two critical Claude Code CVEs enabling RCE and API key theft via malicious repo files — triggered before trust dialogs. Now patched.
Claude Is Dethroning ChatGPT in App Stores as 1M+ Sign Up Daily Amid Pentagon Dispute
Claude surpasses ChatGPT on US App Store with 1M+ daily signups and revenue jumping to $19B annualized — a direct result of the Pentagon supply-chain dispute.
Nano Labs Launches iPollo ClawPC A1 Mini — Dedicated Hardware for OpenClaw Ecosystem
Nano Labs debuts the iPollo ClawPC A1 Mini — the first purpose-built hardware product for the OpenClaw AI agent ecosystem, running LLMs locally on compact hardware.
OpenAI Launches GPT-5.4 With Native Computer-Use Capabilities and 1M Token Context
OpenAI's GPT-5.4 ships with native computer-use, 1M-token context, and 47% fewer tokens on MCP benchmarks — a direct upgrade for autonomous agent pipelines.
Xiaomi Begins Closed Beta of miclaw — Mobile AI Agent Built on MiMo That Can Operate Your Smartphone
Xiaomi's miclaw enters closed beta — a MiMo-powered mobile AI agent that controls apps, taps buttons, and manages smart devices autonomously on Android.
ExpressVPN Launches Industry-First MCP Server — AI Agents Can Now Control VPN Connections
ExpressVPN's new MCP server beta lets AI agents read and control VPN connections from inside code editors — an industry first.
Mastercard Launches 'Verifiable Intent' — Open Standard to Authorize AI Agent Transactions
Mastercard's 'Verifiable Intent' framework lets AI agents make purchases with full authorization, intent verification, and compliance traceability.
OpenAI Open-Sources Symphony — Elixir-Based Agentic Framework for Autonomous Coding Project Management
OpenAI's Symphony is an Elixir-based open-source framework connecting issue trackers to autonomous coding agents via structured implementation runs.
Pentagon Formally Designates Anthropic 'Supply-Chain Risk to National Security' — What's Changed Since Our Last Coverage
The Pentagon formally notified Anthropic of its supply-chain risk designation on Mar 5 — a legal escalation as both sides reportedly return to negotiations.
Anthropic Cowork Preview: Claude as Active Desktop Agent for Windows
Anthropic's Cowork turns Claude into a native Windows desktop agent — file management, automation, and business plugins now in preview.
AWS Officially Adds OpenClaw to Amazon Lightsail as One-Click Blueprint
AWS officially launches OpenClaw on Lightsail — one-click private AI agent deployments with Bedrock, sandboxing, and HTTPS in 15 regions.
Malware-Laced OpenClaw Installers Get Bing AI Search Boost
Fake OpenClaw installers on GitHub were boosted by Bing AI search, delivering GhostSocks malware to unsuspecting users — a critical safety warning.
OpenClaw Agent Based on Anthropic Claude Opus Almost Gets a Job
An OpenClaw agent autonomously created online identities and nearly passed a job application — a landmark moment for agentic AI autonomy and ethics.
OpenClaw Overtakes React and Linux as GitHub's Most-Starred Project — 250k Stars in Weeks
OpenClaw becomes GitHub's most-starred project ever, surpassing React and Linux at 250k+ stars — but security experts are raising enterprise risk alarms.
Defense Contractors Are Dropping Claude After Pentagon's Anthropic Blacklist
Defense tech companies are dropping Claude as the Pentagon's Anthropic blacklist creates impossible contract compliance choices for government contractors.
Zenity Discloses PerplexedAgent: Calendar Invite Hijacks Perplexity Comet Browser, Steals Credentials
Zenity discloses PerplexedAgent: a calendar invite could silently hijack Perplexity Comet, steal credentials, and access local files — patched but category risks remain.
Guild.ai Raises $44M to Help Companies Deploy AI Agents at Scale — Khosla Leads Round
Guild.ai raises $44M at $300M valuation in a Khosla-led round to help enterprises build and deploy AI agents at scale — part of today's agentic infra funding surge.
Klarna and Stripe Launch Shared Payment Tokens — AI Agents Can Now Checkout on Your Behalf
Klarna and Stripe launch Shared Payment Tokens enabling AI agents to complete BNPL purchases on behalf of users — a landmark for agentic commerce infrastructure.
Critical CVE in MS-Agent AI Framework Allows Full System Compromise via Agent Hijacking
CVE-2026-2256 in ModelScope's MS-Agent allows full system compromise via prompt injection — affects v1.6.0rc1 and earlier, patch urgently needed.