zhaoyue722/llm-usage-mcp

What it does

zhaoyue722/llm-usage-mcp is an MCP (Model Context Protocol) server that tracks and analyzes LLM API costs in real-time. It captures usage data from multiple providers including Anthropic, OpenAI, Qwen, and DeepSeek, maintaining a local SQLite database ledger. The tool provides spend queries, detailed provider comparisons, and actionable cost optimization recommendations with first-class support for Chinese providers and CNY to USD currency conversion.

How to set up

Installation requires Python 3.7+ and pip. Clone the repository from GitHub, install dependencies via pip, configure your API keys for supported providers in environment variables, and initialize the local SQLite database. The MCP server can then be registered with compatible clients like Claude Desktop or other MCP-compatible applications to expose cost tracking as integrated tools.

Key features

• Multi-provider API cost tracking (Anthropic, OpenAI, Qwen, DeepSeek)
• Local SQLite database for privacy and control
• Spend analysis and historical cost queries
• Provider comparison reports and benchmarking
• Cost optimization recommendations
• Chinese provider support with CNY/USD conversion
• MCP tool integration for seamless client embedding