AI Coding Tool Pricing: Copilot, Codex and the $20 Asterisk

AI coding tools now come with a headline price and a meter. Engineers are learning to read the asterisks like a taxi fare.

$20/Month* (*Until About Tuesday)

Satire about real events: the jokes are ours, the facts are sourced below.

The pricing page says $20 a month. Somewhere below it, in type you need a magnifying glass for, are the real terms. Anyone who has run an agent for a full workday knows the number at the top is where the bill starts, not where it ends.

That is the joke on our latest design, and this year it has been getting less like a joke and more like a line item.

What happened

GitHub moved Copilot to metered AI Credits on June 1, 2026. Autocomplete stays unmetered. Agent mode and premium models draw down credits at $0.01 each. OpenAI moved Codex to token credits in April. Cursor and Claude Code mix seat fees with usage pools, rate-limit windows or overage.

The entry price is a floor. According to Developers Digest and CloudZero, the $10 to $20 sticker price is where plans start. Daily agent use is commonly quoted at $60 to $200 a month, and heavy users go well past that.

The cost of all those tokens has also reached the business pages. Anthropic's amended S-1 on Nov 2, covered by CNBC, put the scale of compute spending in the headlines.

Companies are reacting with spend caps, token budgets, leaderboards and FinOps-for-tokens efforts. TechCrunch and The Pragmatic Engineer cover the scramble. Reports say Microsoft set internal token budget targets and Uber used up its 2026 AI coding budget by April. MIT Technology Review argues in its headline for counting what ships instead of tokens.

Why engineers get it

There is no shared unit. One tool bills in credits, another in tokens, another in 5-hour windows, weekly caps or premium request pools. A team running three agents gets three partial bills, each in a different currency, and none of them reconcile with the others.

Then there are the classic ways to burn a budget:

None of these feels expensive while you are doing it. The work just keeps going, and the meter keeps running in a tab you are not looking at.

So engineers have picked up a new habit: glancing at spend the way you glance at a taxi meter. You check it at the start, you check it halfway, and you try not to think about it during the long red light of a stuck agent.

Plan lineups keep shifting too. The tier you budgeted for in spring may be a different shape by autumn. Reading the footnotes is now part of picking a tool, right next to reading the docs.

Nobody has to be the villain here. Usage-based billing is a reasonable way to price something whose cost scales with use. The pain is in the gap between a flat, friendly headline number and a variable, multi-unit reality. That gap is where the asterisks live.

The design

The design is a pricing-page card with a huge $20/mo and a tiny asterisk. Under a magnifying glass, the microscopic footnote reads "until about Tuesday". Beside it, a checklist ticks "unlimited*", "fast*" and "agents*", each with an asterisk of its own. The printed line reads "$20/month* *until about Tuesday".

It works on a mug for the person who has just opened the usage dashboard, and on a tee for the person who has to explain it to finance.

Sources

  1. AI Coding Tools Pricing Comparison 2026 - Developers Digest
  2. AI coding tools cost compared: Copilot, Cursor, Claude Code & Codex - CloudZero
  3. The token bill comes due: Inside the industry scramble to manage AI's runaway costs - TechCrunch
  4. Anthropic reports $17 billion in Q3 revenue, purchase commitments of $895.4 billion in revised S-1 - CNBC
  5. The Pulse: token spend breaks budgets - what next? - The Pragmatic Engineer
  6. Stop counting tokens, start counting what ships - MIT Technology Review

More designs