惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

Blog — PlanetScale
Blog — PlanetScale
Vercel News
Vercel News
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
量子位
Y
Y Combinator Blog
IT之家
IT之家
博客园 - 聂微东
L
LangChain Blog
爱范儿
爱范儿
H
Help Net Security
GbyAI
GbyAI
F
Fortinet All Blogs
B
Blog
Microsoft Security Blog
Microsoft Security Blog
罗磊的独立博客
C
Check Point Blog
博客园 - 三生石上(FineUI控件)
小众软件
小众软件
D
DataBreaches.Net
Last Week in AI
Last Week in AI
WordPress大学
WordPress大学
B
Blog RSS Feed
酷 壳 – CoolShell
酷 壳 – CoolShell
宝玉的分享
宝玉的分享

Towards AI

Building AI Agents in Rust — part 4 | Towards AI Building AI Agents in Rust — part 5 | Towards AI The Verified Identity Agent Bridge | Towards AI You Can’t Prompt Your Away Your LLM Problems | Towards AI The Free Agent Trap | Towards AI Your Agentic Loop Will Drift. Here Is the KL Divergence Equation That Measures How Far It Has Wandered From Its Original Instruction. | Towards AI Beyond Chat: Processing Images, PDFs, and Documents with the OpenAI Adapter in Oracle Integration Cloud | Towards AI Building AI Agents in Rust — part 3 | Towards AI Self-Hosting Airflow at Home: Automating Stock Price Data Collection | Towards AI The 76-Hour Frontier: How the Takedown of Claude Fable 5 Birthed the Military-Industrial-AI Complex | Towards AI I Trained a Markdown File to Boost GPT-5.5 by 23 Points — It Shouldn't Work | Towards AI We Replaced ChatGPT With a Local AI Server. Six Months of Honest Data. | Towards AI What Really Makes Cars Pollute? A Data Science Deep Dive into CO₂ Emissions | Towards AI Training GPT-2 From Scratch on a GTX1050 | Towards AI Principal Component Analysis (PCA): Theory, Mathematics, and Applications Build a Zero-Cost Web Automation Pipeline With OpenRouter, OpenClaw, and MediaUse I Gave Qwen3.7-Plus a Screenshot and It Found the Exact Pixel to Click for $0.40 Beyond the Prompt: Why Autonomous AI Agents Are Replacing the Chatbot Moonshot Cracked Claude Code’s Playbook with an MIT Terminal Agent and a $0.60 Model Connections, Roles, and Warehouses: Getting CoCo Desktop Production-Ready from Day One My First $5,000 Month Writing About AI Engineering on Medium Google Shrank Gemma 4 by 72% and Unsloth Fixed the 4-Bit Bug Nobody Else Caught on One 4090, and 4-Bit Shouldn’t Be This Good LangChain Explained: Understanding Models, Prompts, Chains, Memory, Indexes, and Agents TOON: Beyond JSON for LLMs Claude Code Casual, Pro, Elite: The Three Working Personas of Claude Code Mastery MiniMax M3 Decodes 1M Tokens 15x Faster — and It Shouldn’t Be This Cheap Using Amazon SQS for AI Agent Orchestration I Ran a 1.5B-Active Model on My Laptop That Embarrassed a 26B by 46 Points How to Build a Self-Improving Company with AI Part 3 — Implementation/Engine-Level: Choosing the Runtime That Gives You These for Free
Building Production MCP Servers: What the Spec Won’t Tell...
Editorial Team · 2026-05-26 · via Towards AI

Free: 6-day Agentic AI Engineering Email Guide.
Learnings from Towards AI's hands-on work with real clients.

Building Production MCP Servers: What the Spec Won’t Tell You

Last Updated on May 27, 2026 by

Author(s): Can Demir

Originally published on Towards AI.

The token bill, the tool-count cliff, and the “expected behavior” security handoff — three costs MCP delegates without naming them.

Scroll through developer forums this month and you’ll see the same obituary in every thread: MCP is dead. Eric Holmes’ “MCP is dead, long live the CLI” hit the top of Hacker News. Pieter Levels called MCP “just as useless of an idea as LLMs.txt.” Thoughtworks put “naive API-to-MCP conversion” in the Hold ring of their Tech Radar.

Building Production MCP Servers: What the Spec Won’t Tell You

After the lead, the article argues that MCP isn’t really a protocol so much as a cost model disguised as JSON-RPC: each turn forces the model to pay a token “manifest tax” for every tool, large tool lists create a discontinuous “tool-count cliff” where tool selection collapses due to attention/selection saturation, and—per Ox Security—STDIO execution pushes sanitization risk to developers as an “expected behavior,” enabling command injection unless you sanitize before launch. It also highlights an operational failure mode (“stdout corruption”) where stray logs/prints break JSON-RPC framing. Finally, it lays out practical design guidance (limit tools, use progressive disclosure, sanitize at every layer you control, log to stderr only), when MCP is worth the overhead (multi-user/enterprise governance), and what remains uncertain (whether progressive disclosure becomes default, how Anthropic’s stance evolves, and whether model-tool selection cliffs are intrinsic or training artifacts).

Read the full blog for free on Medium.

Published via Towards AI


Towards AI Academy

We Build Enterprise-Grade AI. We'll Teach You to Master It Too.

15 engineers. 100,000+ students. Towards AI Academy teaches what actually survives production.

Start free — no commitment:

6-Day Agentic AI Engineering Email Guide — one practical lesson per day

Agents Architecture Cheatsheet — 3 years of architecture decisions in 6 pages

Our courses:

AI Engineering Certification — 90+ lessons from project selection to deployed product. The most comprehensive practical LLM course out there.

Agent Engineering Course — Hands on with production agent architectures, memory, routing, and eval frameworks — built from real enterprise engagements.

AI for Work — Understand, evaluate, and apply AI for complex work tasks.

Note: Article content contains the views of the contributing authors and not Towards AI.