AgentCostLens — See the real cost of every coding agent task

Track token and dollar spend per coding agent task, so automation never costs more than a human

Problem

A post questions whether AI actually solves problems when token and processing bills often exceed hiring a human. Developers juggling multiple agents (Claude Code, pi, etc.) have little visibility into per-task cost or how much delegating to cheaper models could save.

Solution

A local wrapper that intercepts and logs token usage and model costs for each agent session, broken down by task, project, and agent type. Supports budget alerts and estimates savings from switching to cheaper models.

Users

Indie developers, small teams, and tech managers who heavily use multiple coding agents, especially cost-sensitive ones

MVP

Wrap common agent CLI entry points to parse token usage and model pricing; generate per-task cost reports; support budget alerts; estimate savings from cheaper models; export CSV

Revenue

Free basic stats; subscription for team dashboards, budget policies, and auto-delegation suggestions

Why now

Coding agent usage is exploding but cost visibility is near zero. A post already tried delegating Claude Code work to a cheaper model for 33% savings, proving real demand.