AI CODING PLAN COST CALCULATOR · RUNS IN YOUR BROWSER

Is your AI coding plan cheaper than the API — or is the API cheaper than your plan?

Vendor pricing pages quote a flat monthly price on one page and a per-token API rate on another, and never do the multiplication for how you actually use the tool. This does — enter your plan, your usage, and the model's rates (or start from a preset and edit it), and it works out both monthly costs and the usage level where they cross.

Nothing is uploaded. Every number here stays in this browser.

1. Your subscription

Leave the allowance blank and this compares a flat fee against the API, which is the right shape only if your plan never cuts you off. If yours does, put your own estimate of the monthly token allowance in — usage past it is priced at the same API rates as the other column, and the result will tell you which day of the month you run out. There is no default here on purpose: vendors publish these limits in messages, “premium requests”, or rolling five-hour windows, never in tokens, so any figure we pre-filled would be invented rather than sourced.

2. Your usage (per active day)

Cache reads usually dwarf everything else in an agentic coding session — most of the context is the same files and history sent again, billed at a fraction of the fresh-input rate. If you don't know your numbers, pick a starting point above and adjust it.

3. API rates it would cost to run that usage yourself

Picking a plan above also switches this to that vendor's own API rates — Claude for Claude and Cursor plans (Cursor runs several models; Claude is the closest single default, and Gemini is also here as a preset if that's what you actually run through it), OpenAI's API for ChatGPT Plus and GitHub Copilot. Anthropic charges extra to write to its prompt cache and less to read it back; OpenAI and Google do not bill the write step at all, only a discounted read, which the two new presets model as a $0 cache-write rate.

Why this differs from a simple "tokens × price" calculator

Most cost calculators multiply input and output tokens by a blended rate and stop. That understates the case for a subscription for exactly the users asking this question: agentic coding tools resend most of the file and conversation context on almost every turn, and providers that support prompt caching bill that repeated, unchanged prefix at a small fraction of the fresh-input rate. See how AI coding plan pricing actually works for the full breakdown, sourced examples, and the vendor pricing pages to check before you decide, or this calculator vs. doing that comparison by hand for why the arithmetic is worth automating.

Already seen a blog post work this out by hand for their own account? See this calculator vs. doing the math yourself for why their numbers aren't yours.

Building or watching MCP servers and AI-agent supply chains too? mcp-supply-chain-watch monitors the servers your agent is already connected to for a change in what they can do.

Comparing raw API usage across vendors instead of one plan against its own API? llm-api-pricing-calculator compares monthly cost across Claude, GPT, Gemini, or a custom provider for a token workload you supply.

Questions

Is Claude Max actually worth it over the API?

It depends entirely on how much you use it and how much of that usage is cache reads, which are billed far below fresh input on providers that support prompt caching. This calculator takes your own token estimates (or a light/typical/heavy starting point you edit) and the plan's price, and reports which is cheaper this month and at what usage level the other one would catch up.

Why does this ask for cache read and cache write tokens separately?

Agentic coding tools resend most of the conversation and file context on nearly every turn. Providers that support prompt caching bill a repeated, unchanged prefix at a small fraction of the fresh-input rate, and publicly reported usage breakdowns for heavy coding-agent use put cache reads above 90% of total tokens. A calculator that only multiplies input and output tokens by a blended rate significantly overstates the API-equivalent cost for exactly the heavy users asking this question.

Where do the default prices and rates come from?

They are list prices and published API rates as of the date shown on the page, recorded here because this tool makes no network call. Vendors change pricing without notice and this calculator cannot know when that happens — every field is a plain number you can edit, and the result is only as current as what you put in. Check the vendor's own pricing page before deciding.

What if my plan has a usage limit?

Put your own estimate of the monthly token allowance in the optional field and the comparison stops assuming the plan is unlimited: usage past the allowance is priced at the same API rates as the other column, the result names the active day you run out on, and an all-in row adds the overflow to the flat fee. There is deliberately no default, because vendors publish coding-plan limits in messages, premium requests, or rolling five-hour windows rather than in tokens, so any figure pre-filled here would be invented rather than sourced. Leave it blank and the plan is treated as unlimited, which is the right shape only if yours never cuts you off.

Is anything I type here uploaded?

No. All arithmetic happens in your own browser. There is no account, no server call, and no analytics vendor.