Codex Usage Reports

Know what every Codex task and turn used

2 followers

Know what every Codex task and turn used

2 followers

Automatic local token reports for Codex. See each task's totals, per-turn usage, observed models, reasoning effort, and subagents. Open source, with no API key or hosted dashboard. Missing data stays explicit. Export HTML, Markdown, and JSON receipts.

Free

Launch tags:Productivity•Developer Tools•Artificial Intelligence

Launch Team / Built With

Customer.ioAutomate Messaging Everywhere — Startups Get 12 Months Free

Promoted

OpenAI

GPT-6 Astra Challenge

What became possible in your product with Astra that was not practical before?

Maker

Astra made it practical for me to turn an internal Codex reporting feature into a standalone open-source plugin and CLI, while checking the accounting details that a simple token dashboard can hide. I used GPT-6 Astra in Codex to work across extraction, implementation, tests, packaging, documentation, and release fixes. The result separates task totals, turn deltas, observed model settings, subagents, and account quota. It handles mismatched native counter sources and late completion records without counting the next turn or rewriting original receipts. The project includes nine card languages, four README languages, synthetic demos, and isolated-install validation. The gain was making that cross-cutting engineering and verification work manageable for one maintainer. Astra helped build the tool; usage calculation itself is deterministic local Python, with no LLM calls, reporting API key, or hosted analytics service.

Report

23d ago

Maker

📌

Hi Product Hunt! I'm the maker behind EasyVibeCoding. Long Codex tasks can span many turns, model changes, and subagents. I wanted to answer a simple question: what did this turn actually use? For the GPT-6 Astra Challenge, I'm sharing Codex Usage Reports: an open-source plugin and CLI that turns native Codex observations into compact local reports. It separates task totals from turn deltas, preserves the model and reasoning effort observed at the time, and shows subagent usage and account quota separately. Counter resets and missing evidence stay partial or unknown. The tricky part is the completion boundary. The inline card is a pre-final snapshot; a bounded local Python worker can reconcile the saved receipt when native completion arrives, without counting the next turn. Original receipts are preserved. The runtime uses Python's standard library, needs no reporting API key, and sends no reports to a hosted analytics service. It's free under MIT, with cards in nine languages. The gallery uses synthetic data rendered by the real app. Install instructions and source are on GitHub. These are observed usage reports, not invoices. I'd love feedback on which part of your Codex usage is hardest to understand.

Report

28d ago

Reviews

OpenAI

GPT-6 Astra Challenge

What became possible in your product with Astra that was not practical before?

Maker

Report

23d ago

Maker

📌

Report

28d ago