In this article
- Quick summary
- How Claude Code pricing actually works
- The access-path cost matrix
- Is there a free Claude Code plan?
- Claude Code on Pro ($20/mo)
- Claude Code on Max ($100 and $200/mo)
- Team and Enterprise pricing
- API pricing (pay as you go)
- Subscription vs API: which is cheaper?
- How to get Claude Code cheaper
- Usage limits you need to know
- The bottom line
- FAQ
Claude Code pricing (Nov 2026): is it free, and what Pro, Max, Ultra and API access actually cost
The short version that almost every other page buries: there is no standalone Claude Code price and no free Claude Code plan. Anthropic bundles Claude Code into its paid chat subscriptions, so the cheapest legitimate way in is Pro at $20/month. From there you can scale up to Max ($100/month) or Ultra ($200/month, the renamed Max 20x), buy Team seats (from $20/seat), negotiate Enterprise ($20/seat plus usage, 20-seat minimum), or skip subscriptions entirely and meter Claude Code per token through the API. This guide lays out every access path, the real cost of each, and the break-even point where the API stops being cheaper than just paying for Pro.
| Pro | Max | Ultra | |
|---|---|---|---|
| Price | $20 monthly, $17 a month annual | $100 a month | $200 a month |
| Usage basis | more than Free, every 5 hours and every week | 5x Pro, every 5 hours and every week | 10x Pro every week; no 5-hour limit if verified |
| Limits rule | 5-hour and weekly limits | 5-hour and weekly limits, plus priority access at high-traffic times | weekly limit; everything in Max |
| Who it is for | our reading: daily coding | our reading: heavy users | our reading: the heaviest users |
| Last verified | 2026-11-18 | 2026-11-18 | 2026-11-18 |
Nesyona is independent of Anthropic and not endorsed by it; figures are read from the vendor's own pages on the date shown.
Desk setup for working with AI tools all day
As an Amazon Associate I earn from qualifying purchases. This page carries a labelled Amazon shelf that earns commission.
Cheapest path
Pro at $20 per month is the cheapest legitimate path that includes Claude Code.
Monthly price ladder
Ultra (formerly Max 20x) at $200, Max at $100, and Pro at $20 per month.
Access paths
Every paid path bundles Claude Code; the $0 Free chat plan does not.
What is the quick summary of Claude Code pricing?
Claude Code pricing is the set of access paths and per-path costs for running Anthropic's terminal coding agent, and the key fact is that none of them is a dedicated "Claude Code" subscription. The product is delivered two ways. First, it is bundled into every paid Claude subscription, so the price you pay is the price of the plan. Second, it can be metered per token by pointing it at an API key. There is no $0 tier and no one-off purchase.
Here is the floor-to-ceiling shape, all figures verified 2026-11-18verified 2026-11-18 against Anthropic's published pricing:
- Free chat plan ($0): does not include Claude Code (the Claude Code cell is blank in the vendor's comparison, read Nov 18, 2026).
- Pro ($20/month): the cheapest path that includes Claude Code.
- Max ($100/month, formerly Max 5x) and Ultra ($200/month, formerly Max 20x): same features as Pro, far more usage (Max is 5x Pro usage; Ultra is 10x Pro usage every week).
- Team (from $20/seat/month annually): Claude Code included with every seat, 2-seat minimum.
- Enterprise (custom): seat-based or usage-based, negotiated with sales (page claim, not re-read Nov 18, 2026). Updated Nov 18, 2026: the Enterprise tab lists $20 per seat per month billed annually, a 20-seat minimum, plus usage at API rates.
- API (pay per token): from $0.10 in / $0.50 out per million tokens on Haiku 5.5 (prompts up to 100,000 tokens) up to $4 in / $20 out on Opus 5.5 (Fable 5.2, a tier above Opus, is $8 / $40; Haiku 4.5 at $1 / $5 is now legacy).
How does Claude Code pricing actually work?
Claude Code pricing works on a two-track model: subscription-bundled usage or per-token API metering, and you pick exactly one per session. On the subscription track, Claude Code is a feature of your plan rather than a separate charge. Your Claude Code work draws from the same session-based usage pool as your chat usage, governed by rolling time windows, not by a meter that ticks up with every token. When you hit a wall, you wait for the window to reset.
On the API track, Claude Code authenticates against an ANTHROPIC_API_KEY and bills per token at standard model rates, exactly like any other application built on the LLM API. Every file it reads, every diff it writes, and every reasoning step it takes consumes tokens you pay for. There is no time-window cap; there is only your bill. Anthropic documents both routes in its Pro and Max guide and its developer pricing docs.
This is the single most misunderstood thing about Claude Code cost. People search for "Claude Code price" expecting a number like "$15/month for the coding tool," and there isn't one. You are choosing between a flat subscription that caps your usage by time, or a metered key that caps nothing but your spending.
You are choosing between a flat subscription that caps your usage by time, or a metered key that caps nothing but your spending.How Claude Code pricing works
What does the full Claude Code access-path cost matrix look like?
The access-path cost matrix is the side-by-side of every way to pay for Claude Code, what each one costs, and how usage is governed. This is the table the rest of the article unpacks. Every figure is verified 2026-11-18verified 2026-11-18 against claude.com/pricing; confirm at publish time because Anthropic revises these.
| Access path | Price | How usage is metered | Claude Code included? |
|---|---|---|---|
| Free chat plan | $0 | Limited, every 5 hours and every week | No (blank in the vendor's comparison) |
| Pro | $20/mo ($200/yr up front) | 5-hour session window + weekly limit | Yes |
| Max (formerly Max 5x) | $100/mo | 5x Pro, every 5 hours and every week (dual weekly limits: page claim, not re-read Nov 18, 2026) | Yes |
| Ultra (formerly Max 20x) | $200/mo | 10x Pro every week; no 5-hour limit for verified accounts (updated Nov 18, 2026; the old row said 20x per session) | Yes |
| Team (standard seat) | $20/seat/mo annual ($25 monthly) | More per seat than Pro, 2-seat min | Yes |
| Team (premium seat) | $100/seat/mo annual ($125 monthly) | Higher than standard seat | Yes |
| Enterprise | $20/seat/mo annual, 20-seat minimum, plus usage at API rates (updated Nov 18, 2026; volume terms via contact sales (page claim, not re-read Nov 18, 2026)) | Seat-based or usage-based at API rates | Yes |
| API (pay as you go) | Per token (see API section) | Metered per input/output token | Yes (bring your own key) |
The matrix makes the structure obvious: subscriptions cap usage by time and are predictable; the API caps nothing and is variable. Team's standard seat is the same headline price as Pro, but it carries a 2-seat minimum, so the real entry cost is at least $40/month annually for a small team.
Is there a free Claude Code plan?
A free Claude Code plan is a no-cost tier that includes the terminal coding agent, and as of September 2026verified 2026-11-18 it does not exist. The $0 Free chat plan gives you basic Claude access on web and mobile, but it does not bundle Claude Code. Anthropic does not publish a fixed numeric message cap for Free; its Pro support article described Pro as "at least five times the usage per session compared to our free service" (quoted from that article (our Sept 2026 read, not re-read Nov 18, 2026); the pricing page now says Pro has more usage than Free, every 5 hours and every week), which positions Free as the 1x baseline with a meaningfully smaller per-session allowance.
So if "free" is a hard requirement, your only real option is to spin up a developer account, claim the small free starter credit grant, point Claude Code at that key, and use it until the credits are gone. That is enough to evaluate the tool, not to work in it daily. For where the line sits on the consumer side, see our companion guide on whether Claude itself is free.
What do you get with Claude Code on Pro at $20/month?
Claude Code on Pro is the bundled terminal agent included with Anthropic's entry paid plan, and at $20/month billed monthlyverified 2026-11-18 it is the cheapest legitimate way to use the tool. Annual billing brings the effective rate to roughly $17/monthverified 2026-11-18, charged as $200 billed up frontverified 2026-11-18 (that is Anthropic's own rounding of $200 over twelve months).
Pro gives you "at least five times the usage per session" versus Free (the support article's wording (our Sept 2026 read, not re-read Nov 18, 2026)), with a session-based limit that resets every 5 hours, plus a separate weekly limit that resets seven days after a session starts. Claude Code runs under that same unified subscription, so there is no separate API key to manage and no per-token bill. You also get Cowork and early feature access (page claims, not re-read Nov 18, 2026). Updated Nov 18, 2026: the earlier line that listed priority access during high traffic under Pro was corrected, because the pricing page lists priority access at high-traffic times under Max. The one thing Pro does not cover is API usage, which is purchased separately through the Claude Console.
When is Claude Code on Max worth $100 or $200 a month?
Claude Code on Max is the same bundled agent as Pro with a much larger usage ceiling, and it is worth the step up only when Pro's limits actually block you. Anthropic's Max article describes the two tiers in relative terms, not message counts.
Max, formerly Max 5x ($100/monthverified 2026-11-18)
Five times the usage of Pro, every 5 hours and every week (pricing page, read Nov 18, 2026), with Claude Code under one unified subscription. Max also lists $100 a month in Claude Platform API credits and priority access at high-traffic times. The support article described Max as carrying two weekly usage limits, one across all models and one Sonnet-only, both resetting seven days after a session starts (dual-limit detail: page claim, not re-read Nov 18, 2026). This is the tier for a developer who runs Claude Code for hours most days and keeps brushing Pro's ceiling.
Ultra, formerly Max 20x ($200/monthverified 2026-11-18)
Updated Nov 18, 2026: the vendor lists Ultra at 10x Pro usage every week, with no 5-hour limit for verified accounts, $200 a month in API credits, and says "The Max 20x plan is now called the Ultra plan." The older framing of this tier as twenty times Pro's per-session usage with a dual weekly-limit structure is superseded. This is effectively all-day, every-day agentic coding headroom. If you are not already exhausting Max, you do not need it. For that decision see Claude Max vs Ultra.
The honest verdict: most solo developers never need Max. If you want to pressure-test whether the jump pays off before committing, our deeper breakdown of every Claude plan compared walks through the usage-counting method.
How does Claude Code pricing work for Team and Enterprise?
Team and Enterprise pricing are Anthropic's multi-seat paths, and on both, Claude Code is seat-inclusive, meaning every paid seat carries it. Per claude.com/pricing, the standard Team seat is $20/seat/month billed annuallyverified 2026-11-18 or $25/seat/month billed monthlyverified 2026-11-18, for teams of 2 to 150. The premium Team seat runs $100/seat/month annuallyverified 2026-11-18 ($125 monthlyverified 2026-11-18) for higher usage per seat. Anthropic confirms Claude Code is included with Team and Enterprise seats in its Team and Enterprise guide.
Because of the 2-seat floor, the real entry price for a team is at least $40/month annually (2 standard seats), with the per-seat headline matching Pro.
Enterprise is fundamentally custom and negotiated through sales. The pricing page currently surfaces a "$20/seat, usage cost scales with model and task" starting point, with a 20-seat minimum (updated Nov 18, 2026: the Enterprise tab lists $20 per seat per month billed annually, 20 seat minimum, plus usage at API rates), but treat the negotiated parts as illustrative rather than a guaranteed flat rate (page claim, not re-read Nov 18, 2026). Enterprise runs two models: seat-based (included per-seat usage, Claude Code bundled) or usage-based (no per-seat caps, billed at API rates). Custom rate limits, volume discounts, and dedicated support are arranged via claude.com/contact-sales.
What is Claude Code API pricing on a pay-as-you-go basis?
Claude Code API pricing is the per-token cost of running the agent against your own API key instead of a subscription, billed at standard model rates with no time-window cap. You point Claude Code at an ANTHROPIC_API_KEY, and every token it consumes is billed exactly as documented on Anthropic's developer pricing page. Here are the current first-party rates, all verified 2026-11-18verified 2026-11-18:
| Model | Input (per 1M tokens) | Output (per 1M tokens) |
|---|---|---|
| Fable 5.2 (top tier, updated Nov 18, 2026) | $8 | $40 |
| Fable 5.1 and Fable 5 (older) | $10 | $50 |
| Opus 5.5 (current Opus) | $4 | $20 |
| Opus 5 and Opus 4.5 to 4.8 | $5 | $25 |
| Sonnet 5.5 (current Sonnet) and Sonnet 5 | $2 | $10 |
| Haiku 5.5 (current Haiku), prompts up to 100,000 tokens | $0.10 | $0.50 |
| Haiku 5.5, prompts over 100,000 tokens | $0.50 | $2.50 |
| Haiku 4.5 (legacy next to Haiku 5.5) | $1 | $5 |
| Opus 4.1 (retirement note: page claim, not re-read Nov 18, 2026; Bedrock/Google Cloud only) | $15 | $75 |
Two cost levers matter for Claude Code specifically. First, prompt caching is a big saver because agentic coding re-sends the same project context constantly: a cache read on Opus 5.5 costs $0.20 per million tokensverified 2026-11-18, a twentieth of its standard input rate (a tenth on most other models; Fable 5.1 and 5.2 reads cost 0.025x and Sonnet 5.5 reads 0.05x, updated Nov 18, 2026; cache writes cost 1.25x for 5 minutes or 2x for 1 hour). Second, the Batch API runs at 50% off ($2 input / $10 output on Opus 5.5), though batch is for asynchronous jobs, not interactive coding. The vendor table lists the 50 percent batch discount, and the 1.1x US-only inference multiplier for Claude 4.6 and later models, as stacking with the cache multipliers.
One quiet gotcha: Claude 4.7 and later models use a newer tokenizer that produces approximately 30% more tokens for the same text, so a model can be the same headline rate as a predecessor and still cost more in practice (tokenizer note: page claim (our Sept 2026 read, not re-read Nov 18, 2026)). Budget against real token counts, not just the per-million figure.
Subscription
Flat monthly, capped by time
Cheaper for steady daily coding.
API
Per token, capped by spend
Wins for light or scripted use.
Is Claude Code cheaper on a subscription or via the API?
The subscription-vs-API question is a break-even calculation, and for steady daily coding the subscription almost always wins. Here is the actual math, framed against the Opus 5.5 rate of $4 input / $20 output per million tokensverified 2026-11-18.
A single substantial Claude Code session, the agent reading a real codebase, holding context, writing diffs, running and re-reading test output, can easily move on the order of a few hundred thousand tokens of input plus tens of thousands of output. Charge that at Opus 5.5 API rates and you are often looking at $1 to $3 per serious session even with caching, more without it (our Sept 2026 test estimate, not re-measured Nov 18, 2026). A rough cross-check on the vendor's rates: at a 50/50 blend of input and output on Opus 5.5 the blended rate is $12 per million tokens, so $20 buys about 1.67 million tokens at API rates (our arithmetic, added Nov 18, 2026). Do that several times a day and a metered key passes the $20 Pro subscription's value inside a week.
Run a few serious sessions a day and a metered key passes the $20 Pro subscription's value inside a week.Subscription vs API
So the break-even is roughly this: if you run more than a handful of real coding sessions per week, the flat $20 Pro subscription is cheaper and predictable. The API only wins in three cases:
- Light or occasional use: you touch Claude Code a few times a month, well under a subscription's value.
- Automated or scripted pipelines: you are running Claude Code non-interactively in CI or a batch job, where per-token billing and the 50% Batch discount fit.
- Bursty, unpredictable demand: you would rather pay only for what you use than carry a monthly subscription you sometimes leave idle.
There is also a hybrid worth knowing about: on seat-based plans, when your included subscription usage is exhausted, you can opt in to extra usage credits that continue at standard API rates. That consent is explicit, you will not be silently billed, but it lets a Pro or Max session keep going past the cap at metered rates.
How do you get Claude Code cheaper?
Getting Claude Code cheaper means using the few legitimate cost levers Anthropic offers, because there are no coupons, promo codes, or referral discounts for it. The honest list is short:
- Pay Pro annually. The effective rate drops from $20 to about $17/month ($200 up front), saving roughly $40/year for the exact same access.
- Stay on Pro until limits force you up. Max is double or 5x the price for usage volume only. If you are not exhausting Pro, paying for Max is pure waste.
- Use prompt caching on the API. If you do go metered, caching cuts repeated-context input to about a tenth of standard rate, which matters enormously for agentic coding that re-sends project context.
- Use the Batch API for non-interactive jobs. Asynchronous Claude Code runs qualify for the 50% token discount.
- Claim the new-account starter credits to evaluate before you commit a cent (starter-credit note: page claim, not re-read Nov 18, 2026).
- Spend the API credits your plan lists (Max $100, Ultra $200, Team up to $500 pooled, per the pricing page read Nov 18, 2026) before paying API bills separately.
What you will not find: a Claude Code coupon, a student discount, or an affiliate deal. Anthropic runs no promo program for Claude Code (page claim, not re-read Nov 18, 2026), so any site advertising a "Claude Code discount code" is selling you something that does not exist.
Get the Claude Pricing and Plans Cheat Sheet (PDF)
Every Claude plan and the current API rates on one page, plus the gotchas that change what you actually pay. Re-checked November 2026.
What are the Claude Code usage limits you need to know?
Claude Code usage limits are the time-window allowances that govern how much you can run on a subscription before a reset, and they are relative, not numeric. Anthropic deliberately does not publish a fixed message or token count per window, because the real allowance varies with message length, attachments, conversation history, the model selected, and overall demand.
What is confirmed: Pro and Max reset session usage on a 5-hour rolling window. On top of that, Pro carries a separate weekly limit that resets seven days after a session starts, and Max carries two weekly limits (one across all models, one Sonnet-only; dual-limit detail: page claim, not re-read Nov 18, 2026). Updated Nov 18, 2026: the pricing page lists Pro, Max and Free usage as "every 5 hours and every week," and Ultra as 10x Pro usage every week with no 5-hour limit for verified accounts. Claude Code usage on these plans counts against the same pools as chat usage rather than being metered per token.
The practical takeaway: on Pro, long uninterrupted Claude Code sessions are the thing most likely to hit the wall. If that happens several times a week, that is your signal to move to Max, not to argue with a published number that does not exist.
The bottom line: who should pay for Claude Code, and how
The bottom line is a short decision tree. Claude Code is not free and has no standalone price, so the only question is which bundled path fits your usage.
- Just evaluating? Claim a new API account's starter credits, point Claude Code at the key, and kick the tires until they run out. Zero cost, short runway (starter-credit note: page claim, not re-read Nov 18, 2026).
- Coding most days as a solo dev? Pro at $20/month ($17 effective annually) is the answer. It is the cheapest path that includes Claude Code, and it covers the large majority of individual workloads.
- Hitting Pro's limits several times a week? Step to Max at $100/month, and only to Ultra (formerly Max 20x) at $200/month if you exhaust that too. You are buying usage headroom, not features.
- Running a team? Team seats from $20/seat/month annually (2-seat minimum, so $40/month floor) bundle Claude Code per seat with shared admin.
- Light, scripted, or automated use? Skip subscriptions and meter via the API at standard token rates, leaning on prompt caching and the Batch discount.
For most readers landing on this page as a working developer, the honest answer is Pro. Everything above it is a usage-volume decision, and everything below it is an evaluation period, not a way to work for free. Always confirm the live figures at claude.com/pricing before you subscribe, because Anthropic revises these on its own cadence.
If you are weighing Claude Code against other agentic coding tools before you pay, our hands-on Claude Code vs Cursor vs Copilot comparison puts the price-to-value question side by side. And if you are self-employed, remember that a Claude Code subscription used for work is a deductible business expense, which lowers your effective cost.
Standards and research behind this page
Claims about what an AI tool can and cannot do are easiest to check against the standards bodies and research groups that measure these systems rather than against any vendor's own description. The sources below are the primary ones, linked so a reader can verify a claim here without taking this page's word for it.
- NIST AI Risk Management Framework is the US federal reference for evaluating and governing AI systems, including how their capabilities and limits should be characterised.
- NIST AI RMF Knowledge Base sets out the measurable characteristics of a trustworthy AI system, which is the vocabulary these comparisons use.
- Stanford HAI AI Index publishes the annual measured benchmarks on model capability, cost and adoption that vendor claims are checked against.
- National Science Foundation AI programme funds and documents the underlying research these products are built on.
- arXiv cs.AI carries the primary papers behind most model capability claims, usually months before a product mentions them.
- FTC Endorsement Guides governs how a review or recommendation must disclose a material connection, which is why the disclosure on this page exists.
- US Copyright Office states what copyright covers, which decides who owns the output of a generative tool.
- Section 508 defines the federal accessibility requirements a tool must meet to be usable in public-sector work.
This page compares tools and does not endorse one. Capabilities and pricing change frequently, and a reader should confirm current behaviour with the vendor before relying on it.