TokenTelemetry
See what your AI coding agents think, cost and do
TokenTelemetry provides comprehensive performance monitoring for coding agents by analyzing existing local session logs. Key features include:
• Tokens, cost, and step-by-step trace replay
• Visibility into subagents, skills, and MCP calls
• Compatibility with over 13 coding agents
• No SDK, code changes, or cloud accounts required
• Fully local installation and operation
This robust dashboard reads log files already present on your disk, automatically detecting agents like Claude Code, Codex, Cursor, Gemini CLI, and Copilot. It offers an unparalleled view into agent behavior, expenditure, and reasoning, all consolidated into a single interface. The setup is streamlined, requiring just one command to install, ensuring all data remains securely on your machine.
The tool delivers critical insights into autonomous agent operations, including gateway health, scheduled job status, and source breakdowns across numerous platforms. It also provides per-API-call latency, cache-hit percentages, and detailed subagent delegation rendering. Dedicated pages for skills and memory management offer a complete overview of agent capabilities and resource utilization.
Ideal for developers, engineers, and researchers working with coding and autonomous agents, TokenTelemetry offers a transparent, privacy-focused way to understand, optimize, and manage agent performance without external dependencies.