The Dark Side of MCP Servers

The Dark Side of MCP Servers

Author: Demetrios June 23, 2026 Duration: 1:09:59

Sam Partee (CTO & co-founder of Arcade.dev) and Nate Barbettini (Founding Engineer at Arcade.dev) sit down at the MCP Dev Summit to unpack what nobody wants to admit about the Model Context Protocol: the security model is still full of sharp edges. From tool poisoning and prompt injection to why OAuth got bolted onto the spec, this is a builder 's-eye view of where MCP breaks — and how to ship agents safely anyway.

What we get into:🔓 OAuth on MCP — Why the spec adopted OAuth as its authorization standard, and the class of spoofing attacks it shuts down.☠️ Tool poisoning — How a malicious server hides instructions in tool descriptions, and why your agent trusts them by default.🧪 MCP Debugger & ToolBench — Shining a light on the rough edges by grading servers from S-tier to F-tier.🖥️ Sandboxing agents — Giving an agent a shell and a file system without handing over the keys to your machine.📜 Allow lists — Why MCP has client-level allow lists but skills mostly don't — and why that worries them.🔄 The auto-update problem — How skills and servers that silently update become a supply-chain risk ("rug pulls").✅ SOC 2, honestly — Why the controls are voluntary, misunderstood, and actually about best practices.🤖 AI-generated PRs — The new behaviors to watch for as agents start writing and merging code.

If you build agents, ship MCP servers, or are responsible for AI security at your company, this one's for you.

🔗 Links & ResourcesArcade.dev: https://www.arcade.devArcade MCP framework (GitHub): https://github.com/ArcadeAI/arcade-mcpSam Partee (GitHub): https://github.com/sparteeNate Barbettini (LinkedIn): https://www.linkedin.com/in/nbarbettiniMLOps.community: https://mlops.community

⏱️ Timestamps[00:00] Skills, agents, and local context

[08:36] MCP Debugger grades your server

[10:34] Why AI clients are still buggy

[20:54] Why agents shouldn’t always have shell access

[22:44] “I have a spicy take.”

[26:27] “Do not build your own auth.”

[31:14] The “checking someone else’s email” problem

[35:40] “OAuth is the best worst option.”

[43:50] The future of AI entertainment

[46:19] Tool poisoning explained

[50:49] “Trust me, bro,” is not a security solution

[52:45] MCP registries as the App Store model

[1:00:28] AI-generated PRs and speed vs quality

[1:02:37] Why behavior-driven development is coming back

[1:08:11] Have we already reached AGI?


#MCP #AIAgentSecurity #ToolPoisoning


Hosted by Demetrios, MLOps.community is a space for honest, meandering talks about the real work of making artificial intelligence systems actually work. This isn't about hype or theoretical papers; it's about the messy, practical, and often surprising journey of taking models from a notebook into a live environment. You'll hear from engineers and practitioners who are in the trenches, discussing the tools, the frustrations, and the occasional breakthroughs that define the day-to-day. The conversations are deliberately relaxed, covering everything from traditional machine learning pipelines to the new world of large language models and even the intangible "vibes" of team culture and process. Each episode peels back a layer on what "production" really means, whether that involves deploying a predictive service, managing an agentic system, or maintaining reliability as everything scales. Tuning into this podcast feels like grabbing a coffee with colleagues who aren't afraid to dig into the technical nitty-gritty while keeping the tone conversational and accessible. It's for anyone who builds, manages, or is just curious about the operational backbone that allows AI to deliver value, offering a grounded perspective often missing from the broader conversation.
Author: Language: en-us Episodes: 50

Agentic Conversations (formally mlops.community)
Podcast Episodes
We Cut LLM Latency by 70% in Production [not-audio_url] [/not-audio_url]

Duration: 1:05:20
Maher Hanafi is an engineering leader who went from zero AI experience to self-hosting LLMs at enterprise scale — managing GPU costs, optimizing inference with TensorRT LLM, and building an AI platform for HR tech. In th…
Getting Humans Out of the Way: How to Work with Teams of Agents [not-audio_url] [/not-audio_url]

Duration: 50:30
Rob Ennals is the creator of Broomy, an open-source IDE designed for working effectively with many agents in parallel. He previously worked at Meta, Quora, Google Search, and Intel Research. He has a PhD in Computer Scie…
Fixing GPU Starvation in Large-Scale Distributed Training [not-audio_url] [/not-audio_url]

Duration: 52:48
Kashish Mittal is a Staff Software Engineer at Uber, working on large-scale distributed systems and core backend infrastructure.Fixing GPU Starvation in Large-Scale Distributed Training // MLOps Podcast #367 with Kashish…
arrowspace: Vector Spaces and Graph Wiring [not-audio_url] [/not-audio_url]

Duration: 56:01
Lorenzo Moriondo is a Technical Lead for AI at tuned.org.uk, working on AI agent protocols, graph-based search, and production-grade LLM systems.arrowspace: Vector Spaces and Graph Wiring // MLOps Podcast #365 with Loren…
Agentic Marketplace [not-audio_url] [/not-audio_url]

Duration: 51:26
Donné Stevenson is a Machine Learning Engineer at Prosus, working on scalable ML infrastructure and productionizing GenAI systems across portfolio companies.Pedro Chaves is a Data Science Manager at OLX Group, working on…
Durable Execution and Modern Distributed Systems [not-audio_url] [/not-audio_url]

Duration: 1:00:36
Johann Schleier-Smith is the Technical Lead for AI at Temporal Technologies, working on reliable infrastructure for production AI systems and long-running agent workflows. Durable Execution and Modern Distributed Systems…