spanlens/Spanlens

What it does

Spanlens is an LLM observability and monitoring tool designed for teams using large language models at scale. It provides comprehensive insights into LLM usage patterns, costs, and performance metrics across multiple providers including OpenAI, Anthropic, and Google Gemini. The MCP server integration allows seamless querying of observability data from any MCP-compatible client.

Key features

• Request logs and query history tracking
• Agent trace visualization and debugging
• Cost statistics and spending analytics
• Anomaly detection for unusual usage patterns
• Model savings recommendations
• Per-user and per-team analytics
• Support for OpenAI, Anthropic, and Gemini models
• Open source and self-hostable deployment options

How to set up

Install the Spanlens MCP server using the command: npx y @spanlens/mcpserver. Once installed, configure your LLM provider credentials (OpenAI, Anthropic, or Gemini) in the settings. Connect the MCP server to your preferred MCP client (Claude, ChatGPT, or other compatible platforms). The server will begin collecting and analyzing telemetry data from your LLM requests immediately.