MCPs may be more typed, but skills cost more tokens and fail harder on bash
Vjeux · x · 2026-07-23
A discussion compares MCP and skills in agent workflows:
- MCPs are described as more strongly typed and more native-capability driven
- Skills are said to be more token-hungry
- LLMs are still imperfect at writing bash, which can create more errors
The poster argues the debate is basic reasoning about token use and recall, not PhD-level material.
More from coding & agent
- 7 Core Prompting Rules for Claude 3.5 Opus — JafarNajafov · 2026-07-23
- llama.cpp PR would let all AI-generated code into the project — jacek2023 · 2026-07-23
- Creator shares a Codex + ChatCut workflow for turning long videos into platform clips — yihui_indie · 2026-07-23
- WebMCP in Chrome 101 gets a practical how-to article — amitmerchant · 2026-07-23
- Agent graphs fix handoffs, but every node still rediscovers the codebase — Few_Doughnut4293 · 2026-07-23
- Claude Code’s CLAUDE.md is only context; hard rules need PreToolUse hooks — thisdudelikesAI · 2026-07-23