Claude Code doesn't have a price of its own. It comes with every paid Claude plan, from £15 a month, and it shares that plan's usage allowance with chat. The other way to pay is per token through the Claude API, billed in dollars. Anthropic's own figure for real teams is about $13 per developer per active day, or $150 to $250 a month. Which route is cheaper depends on how much your developers use it and whether you need a fixed monthly cost.
Plan prices below are from Anthropic's UK pricing page, checked on 5 November 2026. The page says the prices include 20% VAT. API prices are in US dollars per million tokens, which is how Anthropic bills them.
Two ways to pay for Claude Code
The first is a subscription. Pro, Max, Ultra, Team and Enterprise all include Claude Code. Your terminal sessions and your chats draw from the same pool, which resets every five hours and every week. When you hit the limit you wait, upgrade, or turn on usage credits, which are charged at API rates.
The second is the API, through a Claude Console account. You pay for every token Claude reads and writes. Nothing stops at a limit, so cost tracks usage exactly, in both directions.
Subscription prices in pounds
| Plan | Price | Claude Code usage | Good for |
|---|---|---|---|
| Pro | £15 a month billed annually (£180 a year), or £18 monthly | Base allowance, shared with chat | Trying it out, or light daily use |
| Max | £90 a month | 5x Pro, every 5 hours and every week | One developer using it most of the day |
| Ultra | £180 a month | 10x Pro each week, no 5-hour limit for verified accounts | Long sessions and autonomous runs |
| Team, Standard seat | £18 a seat a month billed annually, or £22.50 monthly | More than Pro | Non-developers on a team plan |
| Team, Premium seat | £90 a seat a month billed annually, or £113 monthly | 5x a Standard seat | Developers on a team plan |
| Enterprise | US$20 a seat a month plus usage at API rates, billed annually, 20-seat minimum | Metered at API rates | Larger firms that need SCIM, audit logs and spend controls |
Max and Ultra also come with monthly API credits, $100 and $200 respectively, and Team plans get up to $500 a month in pooled API credits. Those credits are for building on the Claude Platform, so they help if your team also calls the API from its own code.
API rates per model
| Model | Input | Output | Cached input (read) |
|---|---|---|---|
| Fable 5.1 | $10 | $50 | $0.25 |
| Opus 5.5 | $4 | $20 | $0.20 |
| Sonnet 5.5 | $2 | $10 | $0.10 |
| Haiku 5.5 | $0.10 | $0.50 | $0.01 |
Cached input matters more than it looks. Claude Code re-sends a lot of the same context on every turn, and it caches it automatically, so a large share of what you pay for is billed at the cache rate rather than the full input rate.
Two things push the bill up. Fast mode on Opus 5.5 runs up to 2.5 times faster at twice the standard price. The 50% batch discount doesn't help here, because interactive coding can't wait for a batch to finish.
What a developer actually spends
Anthropic publishes its own numbers. Across enterprise deployments, the average is around $13 per developer per active day and $150 to $250 per developer per month. Nine in ten users stay below $30 on an active day.
Depending on the exchange rate, $150 to $250 a month lands somewhere between the price of Max and Ultra. So a developer who uses Claude Code every working day usually costs about the same either way. The difference is predictability. A subscription stops at its limit. The API keeps going.
The spread is wide for a reason. A small codebase, Sonnet as the default and one session at a time sits at the bottom. A large monorepo, Opus or Fable for everything and several sessions running in parallel sits at the top, and agent teams multiply it again, because each teammate has its own context.
Which option suits which team
One person trying it: start on Pro. If you keep hitting the five-hour limit, that's your answer, and Max is the next step.
One developer using it all day: Max at £90 a month is usually simpler than the API. You know the cost in advance, and you're unlikely to spend much less on tokens.
A team of five to twenty: the Team plan. Premium seats for developers, Standard for everyone else. Five developers on Premium seats cost £450 a month on annual billing. At Anthropic's average figures, the same five on the API would run roughly $750 to $1,250 a month. You also get single sign-on, central billing and no training on your content by default.
Automation and pipelines: the API. If Claude Code runs in CI, reviews pull requests or works unattended, pay per token and set spend limits in the Console. A subscription seat isn't meant for a shared machine.
Twenty or more people with compliance needs: Enterprise, where usage is metered at API rates and admins can set limits per user and per organisation.
Five ways to keep the bill down
Clear between tasks. Stale context is re-read on every message, so /clear when you switch to unrelated work, and /compact with a short note on what to keep when a session gets long.
Pick the model on purpose. Sonnet handles most coding well. Keep Opus for architecture and hard debugging, and give simple subagent jobs to Haiku.
Trim your tools. Switch off MCP servers you aren't using, and prefer command-line tools such as gh or aws, which don't add tool listings to the context.
Write specific prompts. "Fix the failing test in the invoice parser" costs far less than "look at the code and see what's wrong", because Claude reads fewer files to find out what you mean.
Measure before you roll out. Run a pilot with two or three developers for a fortnight, check /usage and the analytics dashboard, then choose seats. Anthropic recommends the same.
Things buyers often miss
Usage is shared. On a subscription, the same allowance covers chat and Claude Code, so a developer who also uses Claude for documents will hit limits sooner.
Enterprise has a floor. The 20-seat minimum means at least $400 a month in seat fees before any usage, so a ten-person firm is better served by Team.
Client data needs its own decision. A coding tool sees whatever is in the repository, including test data and config files. If your code touches client records, read our guide to Claude and UK GDPR for client data before you roll it out.
How we use it
We build Kelriva's own site and internal agents with Claude Code, and Lichia Reghu, our AI engineer, holds Anthropic's Claude Code 101 certificate. If you're rolling it out to a team and want a second opinion on seats, spend limits or what it should be allowed to see, get in touch. Happy to talk it through.
Sources
All checked on 5 November 2026: Claude plans and pricing (UK), Claude Code documentation on costs, Claude API pricing documentation. Prices and plans change, so check Anthropic's pages before you buy.