Tool Poisoning in MCP Servers: The Silent Threat to AI AgentsMCP tool poisoning hides malicious instructions in tool descriptions to hijack AI agents. Learn how MCPSafe's 5-LLM consensus detects it.7 min read·mcpsafe.hashnode.devMay 14
What Is MCP Typosquatting and How to Detect ItMCP typosquatting is a growing supply chain threat. Learn how attackers use lookalike package names to compromise AI agents and how to detect it.6 min read·mcpsafe.hashnode.devMay 13
We Scanned 448 MCP Servers — Here's What We FoundWe scanned 448 MCP servers and found 5,210 vulnerabilities — hardcoded secrets, tool poisoning, typosquatting, and more. Here's what the data shows.5 min read·mcpsafe.hashnode.devMay 12
I Scanned 2,600 Public MCP Servers. The Supply Chain Is Rough.A snapshot of what's broken in the Model Context Protocol ecosystem — from the kind of mistakes you'd expect to a few that should make you pause. I built MCPSafe, a free security scanner for Model Con4 min read·mcpsafe.hashnode.devMay 7