OpenAI Community
2026-08-06 21:20 UTC
By @kelesmert mertkelesed
AI-116-20260806-social-media-5845dfd6
Turnlens, per-turn token usage and cost tracking for Codex
I built an open-source CLI called Turnlens . It shows the token usage and API-equivalent cost of each individual turn while using Codex. Instead of only showing the total usage of a session, it lets you see: token usage and cost for each prompt model and reasoning effort tool calls completed and aborted turns session-level usage reports I originally built it while testing different prompts, skills and reasoning settings, where session totals weren’t enough to understand what was actually using the tokens. To monitor a Codex session: npx turnlens@latest codex To view usage reports: npx turnlens@latest codex report Turnlens runs locally and only reads Codex session files. Nothing from your sessions is uploaded. The displayed cost is an API-equivalent estimate based on published API prices. It is not an additional charge or your ChatGPT/Codex subscription cost. kelesmert/turnlens on github (since i can’t share link) Feedback would be appreciated.
I built an open-source CLI called Turnlens . It shows the token usage and API-equivalent cost of each individual turn while using Codex. Instead of only showing the total usage of a session, it lets you see: token usage and cost for each prompt model and reasoning effort tool calls completed and aborted turns session-level usage reports I originally built it while testing different prompts, skills and reasoning settings, where session totals weren’t enough to understand what was actually using the tokens. To monitor a Codex session: npx turnlens@latest codex To view usage reports: npx turnlens@latest codex report Turnlens runs locally and only reads Codex session files. Nothing from your sessions is uploaded. The displayed cost is an API-equivalent estimate based on published API prices. It is not an additional charge or your ChatGPT/Codex subscription cost. kelesmert/turnlens on github (since i can’t share link) Feedback would be appreciated.
Full article content could not be extracted automatically. Read the original below.
Source:
OpenAI Community
· community.openai.com