Telemetryc: AI Token Spend Analytics
Open-source dashboard that shows engineering teams where their AI token spend goes, per project, pull request, developer, skill and MCP server.
Telemetryc: AI Token Spend Analytics
Role: AI Engineer
Program: Gauntlet AI — Capstone project
Team: Walid Khori, Sebastian Garces, David Aihe and Sandesh Pathak
Live: telemetryc.com
Overview
Token spend is higher than ever, but how it's used, and how efficiently, stays hidden inside the model providers. Their incentive is simple: use as many tokens as possible, as fast as possible. Telemetryc (built as Pellametric, our Gauntlet capstone) is an open-source dashboard that gives engineers and engineering teams that visibility back.
How It Works
- Collect: Pulls the logs on your machine from every AI coding tool and provider, including Claude Code and Codex
- Synthesize: Normalizes sessions, tokens and output into one dataset per developer and per team
- Attribute: Ties spend to the work it produced: projects, pull requests, skills and MCP servers
- Compare: Shows which prompting habits are lean and which waste tokens
Key Features
- Spend per Pull Request: How many tokens went into each PR, and what shipped
- Team View: Delivery and spend broken out by developer
- Skill and MCP Breakdown: Token use per Claude Code skill and per MCP server
- Efficiency Insights: Patterns and trends that show what's working and what isn't
- Open Source and Self-Hostable: Your usage data stays with you
