Real token/cost meter for AI subscription plans
Problem
AI subscription plans (like OpenAI's Plus/Codex tiers) show only opaque percentage-remaining bars with no explanation of how the percentage is calculated, which models consumed what, or how API price cuts translate to plan usage. Users have literally no idea when or how their quota numbers change.
Opportunity
A third-party usage transparency dashboard that reverse-engineers and tracks actual token consumption and per-model cost breakdowns for AI subscription plans, telling users exactly what they're paying for and when they're being throttled.
Market analysis
Validated pain with a crowded free ecosystem on top of it: tokscale, AI Insights, CodexBar-style widgets and several open-source tray apps already parse local agent logs and OAuth quotas for Claude and Codex plans. The durable differentiator nobody owns is the reverse-engineered explanation layer — what the percentage actually means and how it maps to per-model cost.
Market · Power users on ChatGPT Plus/Pro and Codex tiers; HN threads and star counts on the tools show an engaged niche.
Pricing · Comparable tools are free or open source (tokscale, AI Insights); API-side cost dashboards sell B2B seats, but subscription-plan meters have no proven paid price point.
Pros
- + Clear, recurring annoyance for heavy subscription users.
- + Local log parsing works without scraping or credential sharing.
- + Cross-provider view (OpenAI, Anthropic, Gemini) is fragmented across many single-provider tools.
Cons
- − Built on undocumented behavior; OpenAI already killed one third-party usage endpoint before.
- − Strong free alternatives exist (tokscale, AI Insights, multiple OSS widgets).
- − Reverse-engineering quotas risks ToS friction and constant breakage.
Source
Hacker News (comments)
The moat here is negative: the product depends on reverse-engineering quota math that OpenAI can change or hide at any time — it already deprecated the third-party usage endpoint once, forcing every dashboard author to session-token workarounds. The interesting reframing is from “meter” to “explainer”: the percentage bar is not just opaque, it is adversarial by design, blending models with different exchange rates into one number. A tool that models the underlying rate structure and predicts “you will hit the wall at 6pm on this plan” converts a nice-to-have dashboard into something users check daily — which is the only honest path to charging for it.