Claude Code Pricing 2026: What It Costs on Pro, Max, Team and the API
Claude Code is included with a Claude subscription or billed per token on the API. Here is exactly what each route costs, where the free tier ends, and how a startup gets the API side covered.
Claude Code pricing by plan
Claude Code runs in two billing modes. On a subscription, usage draws from your plan allowance and you never see a token bill. On the API, every request is metered at per-token rates. Plan prices are from claude.com/pricing.
| Plan | Price | Claude Code? | How usage is metered | Best for |
|---|---|---|---|---|
| Free | $0 | No | Not available | Chat only |
| Pro | $20/mo ($17 annual) | Yes | Plan allowance, 5-hour and weekly windows shared with chat | Daily use, a few hours of coding |
| Max 5x | $100/mo | Yes | 5x the Pro allowance, higher output limits | Serious daily Claude Code use |
| Max 20x | $200/mo | Yes | 20x the Pro allowance, priority during peak traffic | All-day agentic coding, multiple sessions |
| Team Standard | $25/seat ($20 annual) | Limited | Standard seat allowance, shared with chat and Cowork | Non-engineering team members |
| Team Premium | $125/seat ($100 annual) | Yes | Max-class allowance per seat, admin spend controls | Engineers on a company plan |
| Enterprise | $20/seat + usage | Yes | Seat allowance plus usage credits, org-level spend limits | Larger orgs, compliance needs |
| API (Claude Console) | Per token | Yes | Input/output tokens at model rates, tracked per workspace | Automation, CI, product-embedded usage |
Claude Code is also billable through Amazon Bedrock, Google Cloud's Agent Platform (Vertex AI) and Microsoft Foundry. On those, tokens hit your cloud bill at that provider's Claude rates, which is exactly where AWS Activate and Google Cloud startup credits come in.
Is Claude Code free? Where the free tier actually ends
The short answer is no. The Claude Free plan gives you chat, code generation in the browser, web search, memory, file creation and code execution. It does not give you Claude Code, the terminal agent that reads your repo, edits files and runs commands. To use Claude Code you need Pro or higher, or an API key with credit on it.
There are three routes people mean when they search for a Claude Code free tier:
- New API accounts get a small amount of free credit. Anthropic's docs say new users receive a small allocation to test the API. It is enough to try Claude Code for an afternoon, not to work on it.
- Claude for Startups. If your company was founded in the last four years and has not received Anthropic credits before, you can apply for API credits that Claude Code draws from. This is the real free tier for founders.
- Cloud credits. Run Claude Code against Bedrock or Vertex AI and it bills your AWS or Google Cloud account, which your startup credits already cover.
What you cannot do is get the Pro or Max subscription free. There is no individual student or founder discount on the plans themselves.
Subscription vs API: which is cheaper for you?
Anthropic publishes real numbers for API-billed Claude Code across enterprise deployments: an average of about $13 per developer per active day, $150 to $250 per developer per month, and 90% of users under $30 on any given day. Those numbers make the comparison simple.
One developer
At the $150 to $250 monthly average, Max 5x ($100) is already cheaper than the API and Max 20x ($200) is roughly break-even with far more headroom. Pro ($20) is cheaper still, but its allowance is shared with chat and resets on a rolling five-hour window, so a long refactor can exhaust it by mid-morning. Most full-time users land on a Max plan.
A small team
Team Premium seats at $100 per seat annually give each engineer Max-class usage on one bill, with admin spend limits and the no-training-by-default guarantee. Standard seats at $20 cover the people who mostly chat. Mixing seat types is allowed.
When the API wins
The API is the right choice when Claude Code is not a person at a keyboard: CI pipelines, scheduled tasks, agent teams, or a product that runs Claude Code on behalf of your users. That usage does not fit a per-seat allowance, and it is exactly the usage startup credits are designed to cover.
What happens when you hit a plan limit
On a subscription you see a session limit or weekly limit message with a reset time. You can wait, or turn on usage credits to keep working past the allowance at API rates. Pro and Max users set a monthly spend limit for those credits; Team and Enterprise admins set them per member or per org.
Claude Code API cost by model
If you bill Claude Code through the API, the model you pick sets the rate. Prices per million tokens from Anthropic's pricing docs. Claude Code caches your context aggressively, so most input tokens in a long session bill at the cache-read rate.
| Model | Input / MTok | Output / MTok | Cache read / MTok | Use in Claude Code |
|---|---|---|---|---|
| Claude Sonnet 5 | $2 | $10 | $0.20 | Default for most coding work |
| Claude Opus 5 | $5 | $25 | $0.50 | Architecture, multi-step reasoning |
| Claude Opus 4.8 / 4.7 | $5 | $25 | $0.50 | Previous Opus generations |
| Claude Haiku 4.5 | $1 | $5 | $0.10 | Simple subagent tasks |
| Claude Fable 5.1 | $10 | $50 | $0.25 | Hardest problems only |
Two multipliers to watch. Fast mode on Opus 5 and Opus 4.8 bills at $10 input / $50 output. US-only inference on 4.6+ models adds 1.1x to every token category. Claude Code applies both to the session cost figure in /usage.
How to keep Claude Code costs down
Token costs scale with context. Everything below is from Anthropic's own cost guide and works on both subscription allowances and API bills.
- Clear between tasks. Run
/clearwhen switching to unrelated work. A one-line question in a session that has been open all day still re-reads the whole conversation. - Pick the model for the job. Sonnet handles most coding. Reserve Opus for architecture. Set
model: haikuon simple subagents. - Mind the cache lifetime. Your first message after a break longer than the cache TTL (an hour on a subscription, five minutes on an API key) misses the cache and reprocesses everything.
- Trim MCP servers. Run
/contextto see what is eating space and/mcpto disable servers you are not using. CLI tools likeghandawsare cheaper than MCP equivalents. - Lower thinking effort on simple tasks. Thinking tokens bill as output. Use
/effortto dial it down. - Delegate noisy work to subagents. Test runs and log parsing stay in the subagent's context; only the summary returns.
- Be careful with agent teams. Teams use roughly 7x the tokens of a normal session because each teammate runs its own context window.
Background usage (conversation summaries, status checks) costs under $0.04 per session and is not something to optimize.
For founders
How startups get Claude Code free
There are two routes and most funded startups should use both.
Route one: Claude for Startups. Anthropic's self-serve program gives early-stage companies free Claude API credits, its highest rate limits, and explicit access to Claude Code, the Claude API, Claude Managed Agents and Claude Cowork. You qualify if the company was founded within the last four years and has not received Anthropic startup credits before. Institutional funding or an Anthropic partner VC or accelerator unlocks the larger allocations. The application is a two-minute form at claude.com/programs/startups. Credits apply to the first-party API through the Claude Console only, not Bedrock or Vertex.
The fine print
Route two: cloud credits. Point Claude Code at Amazon Bedrock or Google Cloud's Agent Platform and the tokens land on your cloud bill. AWS Activate credits (up to $100,000, more via accelerators) and Google Cloud for Startups credits (up to $350,000 on the AI track) both cover that spend. This is the route when you have already burned through Anthropic's credits or want one bill for all inference.
Neither route makes the Pro or Max subscription free. Budget one Max or Team Premium seat per engineer for interactive work and put everything automated on credits.
Claude for Startups
Up to $100,000
- $25,000 to $100,000+ in Claude API credits (partner tier)
- Priority rate limits and early access to new models
- Access to Claude Code, Claude API, Managed Agents and Cowork
Step by step: setting up Claude Code on startup credits
Do this in order and the first Claude Code session you run is already on credits.
-
1
Apply for Claude for Startups
Create a Claude Console account on your company domain at platform.claude.com, then submit the form at claude.com/programs/startups with your company email, website and a sentence on what you are building with Claude. Our Anthropic program page lists the eligibility rules and common rejection reasons. -
2
Wait for credits to land in the Console
The credits attach to your Console organization. When you first authenticate Claude Code with that organization, Anthropic auto-creates a workspace called "Claude Code" for centralized tracking. You cannot create API keys for it; it exists only for Claude Code usage. -
3
Install Claude Code and sign in with the Console account
Install per the docs at code.claude.com, runclaude, and choose the Console (API) sign-in rather than a claude.ai subscription. Usage now bills against the credited organization. -
4
Set workspace spend and rate limits
In the Console, set a spend limit on the Claude Code workspace so an unattended agent cannot drain the credits, and a workspace rate limit so Claude Code cannot starve your production API traffic. -
5
Route any cloud-billed usage through AWS or Google credits
For CI, batch jobs or a second team, configure Claude Code for Bedrock or Vertex AI so that spend draws from AWS Activate or Google Cloud credits instead. Track it with OpenTelemetry export, since cloud usage does not appear in Anthropic's dashboards. -
6
Keep a personal subscription for interactive work
Pro or Max for each engineer's day-to-day sessions. It is cheaper than burning credits on human typing speed, and the credits last longer for what they are meant for.
Claude Code cost for a two-engineer startup: the math
Year-one cost for two engineers who use Claude Code daily and also ship a product that calls Claude.
| Setup | Year-one cost | Covers | Gap |
|---|---|---|---|
| API only, no credits | ~$3,600 to $6,000 | Both engineers at Anthropic's $150 to $250/dev/month average | Product usage on top. No spend ceiling. |
| Pro x2 | $408 (annual) | Interactive Claude Code | Allowance runs out fast. Product usage is API billing. |
| Max 20x x2 | $4,800 | Effectively unlimited interactive use | Product and automation still on the API. |
| Team, 2 Premium seats | $2,400 (annual) | Interactive use plus admin controls and SSO | Same gap on product usage. |
| Max 5x x2 + Claude for Startups | $2,400 | Interactive use on subscription, product and automation on credits | Credits expire. Plan the transition to paid API rates. |
The last row is what we recommend. It is the cheapest setup that covers both kinds of usage, and it keeps the credits for the thing they were granted for. See Claude pricing for the full plan comparison and Claude API pricing for worked cost examples.
Questions founders ask
How much does Claude Code cost?
Is Claude Code free?
Does Claude Pro include Claude Code?
Claude Code on Pro vs Max: what is the difference?
How does Claude Code billing work on Team plans?
Can I use AWS or Google Cloud credits for Claude Code?
Does Claude for Startups include Claude Code?
Related guides
Claude pricing
Every plan, and the full API rate table.
Is Claude free?
What the Free plan includes and where it stops.
Claude API pricing
Per-token rates with worked cost examples.
How to get an Anthropic API key
Console setup, free credits and rate tiers.
Claude for Startups application guide
Step-by-step through the credits form.
AWS Activate credits
Cover Claude Code on Bedrock with cloud credits.
