Launched this week

agentrecap

Launched this week

Simple tool to analyze your local coding-agent sessions

9 followers

Simple tool to analyze your local coding-agent sessions

9 followers

Visit website

Simple tool to analyze your local coding-agent sessions. Gives you data beyond cost, including run durations, tool calls made, lines of code touched, messages per thread etc.

Free

Launch tags:Developer Tools•Artificial Intelligence•GitHub

Launch Team

Robynn AI - Self Improving WebsitesWebsites that improve & heal with self-learning

Promoted

Maker

📌

I believe there's a lot of value in quantifying a lot of the other statistics, apart from accuracy, around LLMs and Coding Agents. Currently, a lot in the field is based on "vibes", even including stuff like how long the model takes to do a given task, which is easily quantifiable. Tools that do exist (like ccusage) only calculate cost right now. agentrecap can eventually also calculate stats like "one shot accuracy". Knowing these, and seeing how others are doing can also help people improve their prompting skills. What do you think?

Report

3d ago

Ran it on a few past sessions and the tool-call breakdown surprised me more than the cost numbers did. The lines-of-code touched stat is genuinely useful for spotting when an agent goes off the rails.

Report

18h ago