Track Claude Code and Codex value with agent-roi
agent-roi is a for measuring how much Claude Code or Codex helps during development work. A developer starts a task in the tool, does the work, then stops the task when finished. The tool can then connect AI usage and Git activity from that time window to that specific task.
The goal is to make the value of easier to measure, instead of relying only on a feeling that work was faster. This can help a solo maker or a team decide whether these AI tools are worth the cost and where they are most useful.
Key points
- agent-roi tracks work done with Claude Code or Codex by task.
- The developer manually starts and stops each development task.
- AI usage and Git activity inside that time window are tied to the task.
- The purpose is to estimate the return from more concretely.
- The workflow can support decisions about keeping, expanding, or changing AI tool use.