MCP Servers Exposed, The State of Secrets and How to Fix It
In this Hacker News webinar, Jonathan Sander, Field CTO at Astrix Security, breaks down what Model Context Protocol really is, how MCP servers are being built in the wild, and why credential handling is the make or break for secure AI adoption. He unpacks fresh research across thousands of open-source MCP servers, then shows practical ways to secure access without slowing the business.
Together with James Azar, the host and moderator of The Hacker News webinars, they connect the security realities with the budget, ownership, and time-to-value pressures driving AI projects today
Watch the webinar on-demand to learn:
MCP, explained clearly
What the Model Context Protocol does in enterprise AI, why client-to-server OAuth is not enough, and how downstream services still create risk if you rely on static credentials.
What the data reveals
Findings from analyzing 5,200 open-source MCP server implementations, including how often secrets are hard coded, passed as environment variables, or protected with OAuth, plus where registries and real-world usage diverge.
How to fix it now
A step-by-step approach to inventory MCP servers, eliminate static keys with vault-backed runtime access, enable rotation and auditability, and use Astrix to govern NHIs across AI agents and MCP, including a walk-through of the free open-source MCP Secret Wrapper.