Benchmark of 11 MCP Servers Reveals Token Count Varies Up to 30x by Design

·35 views

A technical analysis published on September 17, 2026, benchmarked 11 widely used Model Context Protocol (MCP) servers to measure how many tokens their tool definitions consume. The study found dramatic variation: Notion required approximately 19,050 tokens while Puppeteer used only around 610, a difference of over 30 times from a single server installation. Two separate measurement studies were compared, and their results for the same servers diverged by as much as 97%, due to differences in tokenizer methods, server versions, and initial configuration states. Both studies noted that token counts carry little meaning without specifying the counting method and exact server version used. The findings suggest that token budget is a fundamental design constraint for MCP tool development, not merely a performance consideration.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

AI Agent Skills Explained: A Complete Overview for Developers

A guide published on DEV Community by Samuel Adekunle on September 18 covers the essential skills behind AI agents. The article is aimed at developers and tech enthusiasts looking to understand how AI agents function. It falls under topics including artificial intelligence, agent-based systems, and practical tech education. The piece is estimated to be a six-minute read and has received community engagement in the form of reactions.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Enterprise AI in 2025: What's Actually Working Beyond the Hype

Despite widespread excitement around artificial intelligence, most enterprises in 2025 are still struggling to move AI projects from pilot to production. Tools like GitHub Copilot and document-processing AI are delivering real but modest gains — typically 20–40% productivity improvements in specific workflows, not the tenfold leaps vendors promise. Retrieval-Augmented Generation (RAG), which combines company data with large language models, is emerging as one of the most practical enterprise applications, improving search and document analysis while reducing false outputs. However, experts warn that hidden costs — including data preparation, infrastructure, and change management — are routinely left out of AI return-on-investment calculations. The broader takeaway is that AI is evolving into core business infrastructure rather than a standalone strategy, and companies that treat it as such are seeing the most measurable results.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

skills.sh Offers Standardised Way to Package and Share AI Agent Instructions

A new open platform called skills.sh, built by Vercel Labs, aims to standardise how AI coding agents receive specialised instructions and workflows. The platform introduces a format called Agent Skills, where each skill is a folder containing a SKILL.md file with YAML metadata and markdown instructions. Developers can install skills with a single command via a CLI tool that is compatible with agents including Claude Code, Cursor, GitHub Copilot, and Gemini CLI. The system uses progressive disclosure, loading only skill names and descriptions at startup to keep context windows efficient, then fetching full instructions when relevant. skills.sh functions as a public directory where users can browse, share, and discover community-built skills across languages and frameworks.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Conversor Brasil launches free browser-based file conversion tool with 115+ features

A Brazilian developer has launched Conversor Brasil, a free file conversion platform that processes all files entirely within the user's browser without uploading data to any server. The tool was built in response to privacy concerns around converting sensitive documents on third-party websites. At launch, the platform offers over 115 tools spanning image conversion, PDF editing, video-to-audio extraction, and data file transformation. It also includes Brazil-specific utilities such as Pix QR code decoding, boleto barcode parsing, CPF/CNPJ validation, and traditional Brazilian unit conversions. The site is built with plain HTML, CSS, and JavaScript, relying on browser APIs like Canvas, File API, Web Workers, and WebAssembly to handle all processing client-side.

0 comments Read more at DEV Community