TokenTelemetry — Local-first observability for AI coding agents: token usage, cost & reasoning traces (like localGPT but for agent monitoring)

[VasiHemanth]

Hey localGPT community!

You're already here because you care about privacy-first, local AI — no data leaving your machine. I built something that shares that same philosophy but for a different problem: observability into your AI coding agent sessions.

Introducing TokenTelemetry — a 100% local, open-source dashboard for tracking what your AI coding agents cost, reason, and do.

The localGPT connection

Just like localGPT keeps your documents on your machine, TokenTelemetry keeps all your agent telemetry local — no cloud, no API calls, no signup. It reads directly from your agent's local log files and renders everything in a local dashboard.

What it tracks

• Token usage & cost per session, per model, in real-time
• Reasoning traces — the full chain of thought your agent used
• Subagent activity — which sub-agents fired, what they did, their cost
• Tool calls & artifacts — every action taken during the session
• 9 agents supported out of the box — zero config

Supported agents

Claude Code, Codex, Gemini CLI, Cursor, GitHub Copilot, Qwen, Vibe, Antigravity, OpenCode — all read from local log files. No proxies, no middlemen.

Install

curl -fsSL https://raw.githubusercontent.com/VasiHemanth/tokentelemetry/main/install.sh | bash

MIT licensed · 100% local · no signup · Node 18+, Python 3.9+

GitHub: https://github.com/VasiHemanth/tokentelemetry
Website: https://tokentelemetry.com

Would love feedback from the local-first AI community here — this feels like exactly the right crowd for this kind of tool!