Claude Code Cost Estimator

Calculate your real monthly Claude Code spend. See if subscription (Pro/Max) or pay-per-token API fits you better.

⚑ Have your actual API logs?

Upload api_costs.json (from ai_client / loggers) β€” auto-detect your usage profile. File parsed locally in browser, never uploaded.

Supported schema
{
  "calls": [
    { "model": "claude-sonnet-4-5", "tokens_in": 1997, "tokens_out": 285, "date": "2026-05-22" },
    { "model": "claude-opus-4-7",   "tokens_in":  443, "tokens_out": 120, "date": "2026-05-22" }
  ]
}

1. Your usage

5Heavy: 100+300

Avg: 20K in / 5K out

All Sonnet (cheap)All Opus (5x cost)
0% (no cache)90% (cached at 10% price)

2. Your monthly API cost

$209880 prompts Β· 17.6M in Β· 4.4M out

πŸ’° Prompt caching saves $52/mo (without cache: $261)

3. Plan cost for you (sorted)

⚠️ Plans' API-equivalent caps are estimates (community-tested + official multipliers). Vendors don't publish real token caps.

Claude Max 5xBestest.
5x Pro usage, heavy Claude Code useOfficial β†’
$100
Cursor Proest.
500 fast premium reqs + unlimited slowOfficial β†’
$169
Sub $20 + over $149
Claude Proest.
~45 msgs / 5h, includes limited Claude CodeOfficial β†’
$179
Sub $20 + over $159
GitHub Copilot Proest.
Unlimited completions + limited chatOfficial β†’
$189
Sub $10 + over $179
Codeium / Windsurfest.
Free unlimited completionsOfficial β†’
$189
Sub $0.00 + over $189
Claude Max 20xest.
20x Pro usage, pro developersOfficial β†’
$200
Claude Freeest.
Free tier, ~30 msgs / 5h, Sonnet onlyOfficial β†’
$204
Sub $0.00 + over $204
Pure API pay-as-you-go
$209

πŸ’‘ Max vs min: $109/mo difference = $1,309/year

πŸ’‘ Recommendation

Based on your usage, best fit is Claude Max 5x at ~$100.

SERVICEΒ·Done estimating?

Need granular AI cost governance for your team?

I run my own AI automation at $20-50/day with full budget caps, kill switches, and fallback chains. From $800, 3-day delivery, source code yours.

Talk to me→Or compare all API prices

FAQ

How much Claude Code does Claude Pro $20/mo include?+
Anthropic doesn't publish real token caps. Community testing suggests ~45 messages per 5h window, roughly $50 API-equivalent with good prompt caching. Heavy users hit limits fast β€” consider Max 5x.
How to choose between Claude Max 5x and 20x?+
Rough API-equivalent estimates: 5x β‰ˆ $250, 20x β‰ˆ $1000. If your API-priced monthly cost is $100-300, pick 5x. Only go 20x above $500. Under $100, Pro is enough.
Which is cheaper: Claude Code or Cursor?+
At $20, Cursor Pro gives 500 fast premium reqs + unlimited slow. Claude Pro is stricter but with Sonnet 4.6/Opus 4.7 advantage. Light use β†’ Cursor Pro. Heavy Opus-dependent work β†’ Claude Max 5x.
Does prompt caching really save 90%?+
Yes β€” cached input tokens are billed at 10% (90% off). But caching only applies to **repeated input** β€” revisiting the same project has 50-70% hit rate; fresh chats hit 0. Heavy Claude Code users typically see 40-70% hit rate.
Should I use Sonnet or Opus?+
Opus is 5x more than Sonnet ($15/$75 vs $3/$15). Daily editing/explain/debug works fine on Sonnet. Reserve Opus for architecture, complex refactors, reasoning-heavy work. 70% Sonnet + 30% Opus is a sane baseline.
When does pure API beat subscriptions?+
When usage is very light (<$20/mo API) or very heavy (>$1000/mo). Light users waste unused subscription. Heavy users past Max 20x get better control with pure API (quota management, multiple keys). Mid-range ($50-500) subscriptions almost always win.
How often is pricing data updated?+
Pricing as of 2026-05-22. I update after vendor announcements, but check official sources (each plan card has "Official β†’" link). Open a GitHub issue if outdated.
What does the 'est.' label mean?+
Anthropic, Cursor, etc don't publish actual token caps per plan. API-equivalent numbers are based on: 1) official multipliers (Max 5x = Pro Γ— 5), 2) community testing. Use as a guide β€” your actual experience may vary.
Which plan should I pick?+
Adjust sliders above to your real usage. Section 3 sorts plans for you β€” best one is highlighted. Rule of thumb: light β†’ Free/Pro, mid β†’ Cursor Pro or Claude Pro, heavy β†’ Max 5x, super-heavy β†’ Max 20x.