CodingPlans

Which AI coding subscription actually gives you the most for your money?

Claude, ChatGPT/Codex, Cursor, GitHub Copilot, Devin/Windsurf, Kiro, Zed and Google Antigravity all sell flat-rate plans, and most of them don't tell you what a dollar of subscription buys. CodingPlans converts every plan into numbers you can compare:

• API-equivalent value: what the included usage would cost at list API prices, and the resulting Value × (API value ÷ monthly price).
• Output tokens per month: that dollar value converted into tokens using a fixed agentic-coding workload (heavy on cache reads).
• Intelligence: the Artificial Analysis Intelligence Index of the best model you can reach on the plan.
• Composite score (0 to 100): a weighted blend of the three, with weights you can change.

Where a vendor publishes the dollar value of a plan, the number is marked published. Where it doesn't, the number is a clearly labelled estimate. The methodology page spells out exactly which is which.

Pages

Route	What it shows
/	Leaderboard, composite-score and value-multiple charts, price vs API-value scatter, vendor cards
/vendors/[slug]	One product's plan ladder, limits, strengths and caveats, value over time
/models	Intelligence vs cost per task, plus a sortable model table
/value-over-time	Dated before/after changes, the snapshot ledger, API price by model generation, plan-change timeline
/methodology	How every number is produced and how reliable it is

Getting started

Requires Node.js 20 or newer.

git clone https://github.com/OwenKruse/value-tracker.git
cd value-tracker
npm install
npm run dev      # http://localhost:3000

Other scripts:

npm run build      # production build
npm run start      # serve the production build
npm run lint       # eslint
npm run snapshot   # append today's prices and value ranges to src/data/snapshots.json

Built with Next.js 16, React 19, Tailwind CSS 4 and TypeScript. There is no backend or database: all data is static JSON in src/data/.

Data

File	Contents
plans.json	Price, API-value range (low / mid / high), basis (published or estimated), confidence, best model
models.json	API prices per 1M tokens and Artificial Analysis intelligence index
vendors.json	Descriptions, strengths, caveats, source links
history.json	Dated events, before/after reference points, API prices by model generation
aa-leaderboard.json	Artificial Analysis leaderboard rows
snapshots.json	Ledger written by npm run snapshot

The scoring logic lives in src/lib/metrics.ts and the types in src/lib/types.ts.

Contributing

Pull requests are very welcome. Plans and prices change constantly, so this project is only useful if it keeps up, and one person can't watch every vendor. You don't need to be a Next.js expert to help: much of the most valuable work is editing JSON.

Good ways to contribute:

• Fix or update a price or limit. A vendor changed a plan? Update plans.json and link the page that shows it in sourceUrl.
• Improve an estimate. If you have measured usage (for example from ccusage or the Codex usage dashboard), a PR that tightens an estimated range, with your method explained, is exactly what this project needs.
• Add a plan, vendor or model. Add it to plans.json / vendors.json / models.json. Use an existing entry as a template.
• Refresh Artificial Analysis data in models.json and aa-leaderboard.json.
• Report an error. Spotted a wrong number or a broken source link? Open an issue with a link to evidence.
• Improve the app. Accessibility fixes, charts, mobile layout, performance and new views are all fair game.

Ground rules for data changes

• Cite a source. Every price, limit or claim should be traceable to a public page. Put the URL in the entry.
• Be honest about certainty. Use basis: "published" only when the vendor states the value in dollars or in credits with a dollar price. Otherwise use "estimated" and set confidence accordingly.
• Run npm run snapshot after changing prices or values so the history chart gets a new data point.

Submitting a pull request

1. Fork the repo and create a branch from main.
2. Make your change. Keep PRs focused: one concern per PR is easiest to review.
3. Run npm run lint and npm run build and make sure both pass.
4. If you changed anything visible, include a screenshot in the PR description.
5. Open the PR and explain what changed and why. For data changes, link your sources.

Not sure whether an idea fits? Open an issue first and ask. Small, rough PRs are welcome too; we can polish them together.

Notes on third-party content

• Intelligence Index scores and cost-per-task figures come from Artificial Analysis.
• Product logos in public/logos/ are trademarks of their owners and are used only to identify the products being compared. Sources and licences are listed in public/logos/README.md.
• CodingPlans is an independent project and is not affiliated with any of the vendors it compares.

License

MIT