I Made MCP 94% Cheaper (And It Only Took One Command)
Hacker News
February 25, 2026
AI-Generated Deep Dive Summary
Making MCP 94% Cheaper with One Command: A Breakthrough in AI Efficiency
In a groundbreaking development for AI agents, researchers have discovered a cost-effective alternative to using Microsoft’s Cognitive Services (MCP) platform. By converting MCP tools into command-line interfaces (CLIs), developers can slash token usage by up to 94%. This innovation not only reduces costs but also simplifies the way AI agents interact with tools and services.
Traditionally, MCP requires agents to load entire tool catalogs as JSON Schema at session start, consuming大量tokens upfront. For example, a setup with 6 MCP servers and 14 tools each would require loading 84 full schemas—totalling nearly 15,540 tokens. In contrast, CLI-based approaches only load essential information when needed. Instead of preloading all tool details, the agent discovers specifics on demand, significantly reducing initial token usage.
The savings are substantial:
- Session start token usage drops from ~15,540 (MCP) to just ~300 (CLI), a 98% reduction.
- Tool calls also benefit, with MCP requiring ~30 tokens per call and CLI needing only ~6 tokens for execution after initial discovery.
This shift mirrors Anthropic’s Tool Search, which loads a search index instead of full schemas, but CLI offers broader compatibility across models. By using CLIHub—a directory of CLIs for agent tools—developers can generate CLIs from MCPs with a single command. This approach not only cuts costs but also enhances scalability, making it easier
Verticals
techstartups
Originally published on Hacker News on 2/25/2026