Tag
10 articles tagged #MCP.
P.01OpenAI's Agents API public beta puts session orchestration, context compaction, and sandboxed execution behind one call. What it replaces and what it costs.
P.02CVE-2026-59822 lets an attacker fake a Bearer token and skip LiteLLM's MCP auth entirely. Different bug from June's RCE chain, same exposed surface.
P.03WebMCP is a W3C proposal letting a page register JavaScript functions as tools an agent calls in-browser, no server MCP. How it works, and where it breaks.
P.04Only 8.5% of public MCP servers use OAuth, and a honeypot got hit within 48 hours. The checklist for auth, tool scoping, and input handling before you ship.
P.05A critical authorization bypass in n8n-MCP let one tenant read, delete, or destroy another tenant's workflow backups. Who's affected, and what to patch.
P.06Malicious AI skills and poisoned CLAUDE.md files are now a supply chain attack vector. Here is how the ClawdHub incident worked and what to do about it.
P.07EditorPickPrompt engineering is dead. Context engineering, managing system prompts, RAG results, tool outputs, memory, and history, is the skill that matters now in 2026.
P.08EditorPickHow we built production multi-agent systems with the Claude Agent SDK and MCP, covering orchestrator-worker patterns, handoffs, error handling, and tracing.
P.09Anthropic launched domain-specific Claude Cowork plugins for legal, finance, sales, and marketing, with MCP integrations for Slack, Figma, and Salesforce.
P.10Model Context Protocol is the new standard for connecting AI to external tools. Here's a practical guide to building, deploying, and debugging MCP servers — with real code examples from production.