[Ep 023] MCP Servers Deep Dive: Secure, Reusable Tools for AI Agents

 AI agents are powerful—but without the right “tooling layer,” they can be risky, inconsistent, and hard to scale. In this episode, Tim Harting and Gavin Franklin break down what an MCP server is (think: a secure, deterministic API layer for agents), why it exists even when you already have APIs, and how it improves reusability across multiple agents. They also walk through three practical adoption paths—existing MCPs, MCP governance platforms, and DIY MCP servers—plus what leaders should ask vendors in 2026 as MCP becomes a standard integration pathway.Highlights / Takeaways:🧠 Agents reason; MCPs execute. Keep the “brain” non-deterministic and the “tools” deterministic.🔒 Less scope = more safety. Expose only the minimum set of tools your agents truly need.🔁 Build once, reuse everywhere. One MCP server can power multiple agents across your ecosystem.🧩 Three adoption lanes: pre-built MCPs, governance platforms (ex: integration/ETL-style), or DIY.❓ New vendor question: “Are you MCP-capable—and what’s your roadmap?”Call to Action:Take one system in your stack this week and ask “What’s our MCP plan?” Do we use an existing MCP, a governance platform like what we discussed, or do we need to build one? If you need help, reach out to us at Cirrius. Contact details belowWant More: Watch: YoutubeListen: Apple Podcasts \ SpotifyDown the referenced visualsRecommended LinksGavin Franklin on LinkedInTim Harting on LinkedInGreg Banks on LinkedIn Cirrius BlogCirrius SolutionsJason Fowler MusicFor questions or feedback, please contact us at [email protected]