MCP Token Analyzer
A powerful CLI tool to analyze token usage and cost impact of MCP servers
Understand exactly how many tokens your MCP servers add to every API call. Get cost estimates across major AI providers and optimize your context window usage.
Quick Start
# Use the standalone script python3 mcp-analyze.py npx shadcn-ui-mcp-server Key Features
Token Analysis
Analyze any MCP server supporting stdio transport. Get detailed token breakdown by tools, resources, and prompts.
Cost Estimation
Cost estimates for OpenAI, Anthropic, Google, and other providers per API call.
Rich Terminal UI
Beautiful formatted tables and colors. Export to JSON for further analysis.
Pure Python
Works on Linux, macOS, and Windows. No Node.js required.
Why Analyze MCP Token Overhead?
Every MCP server adds tokens to your context. Understanding this overhead helps you make informed decisions about cost and context management.
Optimize MCP Configurations
Identify high-token tools that you might want to exclude from your MCP setup to reduce context overhead.
Budget Planning
Estimate API costs before deploying MCP servers in production. Know the per-request token overhead upfront.
Server Comparison
Compare token overhead across different MCP servers to choose the most efficient option for your use case.
Context Management
Understand how much of your context window MCP consumes, leaving room for actual conversation.
How It Works
Connect to Your MCP Server
Enter your MCP server command (like npx shadcn-ui-mcp-server) or SSE URL for remote servers.
Analyze Tools, Resources & Prompts
The analyzer connects to your server via stdio and fetches all exposed capabilities, then counts tokens for each item.
View Token Counts & Cost Estimates
See total tokens, per-category breakdown, and cost impact across GPT-4o, Claude, and Gemini models.
Optimize Your Configuration
Use the detailed breakdown to identify high-token items and decide which tools to include or exclude.
Built With
Language
Pure Python
Transport
MCP stdio protocol
Platforms
Linux, macOS, Windows
Ready to Analyze Your MCP Servers?
Install with pip or clone the repository to get started.