Show HN: Tarmac – Know what Claude Code will cost before you run it (github.com)

Tarmac has launched a tool designed to provide cost estimates for running prompts in Claude Code, addressing a significant pain point for developers. This innovation allows users to preview a calibrated cost range before executing their AI coding tasks, eliminating the unpredictability of costs which can vary widely from $5 to over $20 per request. By intercepting prompts and leveraging a trained regression model with conformal prediction intervals, Tarmac injects a cost estimate into Claude's context, enabling informed decision-making about whether to proceed with execution. The significance of Tarmac lies in its ability to enhance cost transparency in AI development workflows. Traditional methods for cost estimation fell short, lacking the accuracy and reliability that Tarmac promises, with an overall coverage rate of 81.1% on actual costs falling within predicted ranges. This tool operates locally with no external dependencies, providing a fast, user-friendly installation process and a detailed report comparing estimated versus actual costs post-session. As such, Tarmac represents a meaningful leap toward more budget-conscious AI development practices and could pave the way for similar solutions across other AI platforms in the future.