TokenTelemetry
See what your AI coding agents think, cost and do
Start new thread

TokenTelemetry - See what your AI coding agents think, cost and do

by•
TokenTelemetry reads the session logs your AI coding agents already write on disk and turns them into one local dashboard: tokens, cost, step-by-step trace replay, subagents, skills and MCP calls. Works with 13+ agents: Claude Code, Codex, Cursor, Copilot, Gemini CLI, Antigravity, Grok, OpenCode, Cline, Qwen, Pi and Hermes. No SDK, no code changes, no cloud account. One command to install, everything stays on your machine, and it's open source.

Add a comment

Replies

Best
Maker
📌
Hey PH 👋 I use a lot of AI coding agents: Claude Code, Codex, Cursor, Copilot, Gemini, and a handful of others. Every one of them writes a detailed session log to my disk, and I had never once opened one. TokenTelemetry reads all of them and puts them in a single local dashboard. There's no SDK to install and no cloud account. It finds the logs your agents already write, prices them against current model rates, and shows you tokens, cost, and a step-by-step replay of any session. The number that made me finish it: $853 of my spend came from subagents my sessions spawned, across 177 child sessions that never appeared as a line item anywhere. And the agent that used the most tokens wasn't the one that cost the most money, not even close. It's free and open source, runs entirely on your machine, and installs with one command. I'd genuinely like to know: is cost the reason you'd install this, or is it the trace replay? Happy to answer anything in the comments today.

Really cool, Hemanth! One local dashboard for tokens, cost and step-by-step trace replay across 13+ coding agents is exactly the observability devs need. Huge congrats on the launch!

Maker

 Thanks, Zvonimir! Really appreciate the support. Give it a try with your agents and let me know how it goes after you've used it, would love your feedback on what works and what's missing! 🙏