Skip to main content
dijitul.ai
Pricing

Priced in pounds. Metered per key.

An included monthly allowance, a published overage rate, and a hard stop so a runaway loop cannot surprise you. No seat licences and no dollar invoices. All prices exclude VAT.

Know what you need? Compare the plans. Not sure yet? Tell us what you are building and we will work out the plan.

Plans

Solo

For individual developers

£29 per month

excl. VAT · 14-day free trial

Included
5M units
Then
£6.5 / M units
Residency
UK only (London)
  • 5M units included, then £6.50 per million
  • Inference in AWS Europe (London) only — never leaves the UK
  • UK data processing agreement
  • Reversible PII tokenisation
  • 30-day request logs
  • Email support
Start free trial

14 days free · cancel any time

Studio

For small teams

£99 per month

excl. VAT · 14-day free trial

Included
20M units
Then
£6 / M units
Residency
UK only (London)
  • 20M units included, then £6 per million
  • Inference in AWS Europe (London) only — never leaves the UK
  • Per-key model restrictions and IP allow-listing
  • IP allow-listing
  • Custom redaction dictionary
  • 90-day request logs (EU AI Act ready)
  • Priority email support
Start free trial

14 days free · cancel any time

Assurance

Evidence for your auditor

£249 per month

excl. VAT · 14-day free trial

Included
50M units
Then
£6 / M units
Residency
UK only (London)
  • 50M units included, then £6 per million
  • Per-request residency attestation you can forward to an auditor
  • Full audit export for EU AI Act log retention
  • 12-month log retention
  • Inference in AWS Europe (London) only — never leaves the UK
  • Claude Sonnet 4.6 and Claude Opus 4.6
  • Named support contact
Start free trial

14 days free · cancel any time

Agency

Resell to your own clients

£399 per month

excl. VAT · 14-day free trial

Included
90M units
Then
£5.5 / M units
Residency
UK only (London)
  • 90M units included, then £5.50 per million
  • Inference in AWS Europe (London) only — never leaves the UK
  • Unlimited client sub-accounts
  • Per-client usage exports for rebilling
  • Per-client residency attestations you can forward
  • White-label documentation
  • 90-day request logs
  • Named support contact
Start free trial

14 days free · cancel any time

Regulated

Bespoke

Let's talk

Priced against your volume and residency requirements.

Included
300M units
Then
£5.5 / M units
Residency
UK only (London)
  • UK-only processing guarantee
  • Single-tenant deployment option
  • 12-month audit log retention with SIEM export
  • DPIA support pack
  • Security questionnaire assistance
  • 99.9% uptime SLA
Contact sales

Need a different shape — higher concurrency, a private allow-list, or invoicing on a purchase order? Talk to us.

Tell us what you are building

Pick the closest workload and roughly how much of it you will do. We will work out the units and point you at a plan. Nothing to sign up for, and you can change plan any month.

What are you building?

Roughly units per . Estimates, not a quote. The assistant figure is anchored on real traffic through this gateway; the rest are modelled and deliberately cautious, so they read high rather than low. Your first month's dashboard will tell you the truth, and you can move plan then.

Estimated need

units / month

Token pricing

What each model costs to run

Pounds per one million tokens of each type, for reference. You are charged in units against your plan allowance; these are the underlying rates that define how many units each token consumes.

Price per million tokens by model, in pounds sterling
Model Tier Input Output Cache write Cache read
Claude Sonnet 4.6 claude-sonnet-4-6 UK & EEA £3.30 £16.50 £4.15 £0.33
Claude Opus 4.6 claude-opus-4-6 UK & EEA £5.50 £27.50 £6.90 £0.55

All prices exclude VAT, which is added at the prevailing UK rate. Allowances are in units — one unit is one Claude Sonnet input token; an output token counts as 5 and an Opus token as roughly 1.7, matching what each actually costs. Prompt caching is charged at the cache-write rate on the first call and the cache-read rate on subsequent hits — for repeated system prompts that is usually a substantial saving. See the model reference for context windows and cache minimums.

In every plan

Not sold as an add-on

The controls that make this product worth buying are not upsells. They are on by default at every price point.

UK-hosted inference

Every request runs on AWS Bedrock in London. Never silently downgraded, and there is no wider tier to downgrade to.

PII detection and redaction

Four privacy modes — off, flag, redact, tokenise — configurable per tenant and per key.

No prompt logging by default

Bodies are never written to disk unless you explicitly turn on time-limited debug capture.

Per-key metering

Every request attributed to the key that made it, so cost maps to a team, a project or a client matter.

IP allow-listing

Restrict a key to named source addresses so a leaked key is not usable from anywhere else.

Audit trail

Key creation, rotation, revocation and configuration changes are recorded with actor and timestamp.

Questions

Pricing questions

How is a token counted?

Exactly as Anthropic counts it. Input tokens, output tokens, cache-write tokens and cache-read tokens are metered separately and reported back on every response in the usage object, so your figures and ours reconcile without a support ticket.

What happens when I use up my allowance?

On a metered plan, further usage is charged at the overage rate for that plan and appears on the next invoice. There is also a hard stop: once a tenant runs far enough past its allowance, requests are refused rather than continuing to bill. A runaway retry loop cannot quietly generate a five-figure invoice.

Do the different models cost different amounts?

Yes. The plan allowance is expressed in units, and each model consumes that allowance at its own rate: an Opus token weighs roughly 1.7 units against a Sonnet token, and every output token weighs 5.

Which models can I use?

Every plan is UK-only, which means Claude Sonnet 4.6 and Claude Opus 4.6 — the models AWS can invoke in-region in London. Newer Claude models are only published as cross-region inference profiles, which cannot be pinned to the UK, so we do not offer them on a UK tier rather than quietly routing your data around the EEA.

Is there a minimum term?

Monthly plans are month to month and can be cancelled from the dashboard at any time, effective at the end of the current billing period. Annual and enterprise arrangements are agreed in writing.

How do I pay?

By card or direct debit through Stripe, invoiced in pounds sterling. VAT is added at the prevailing UK rate. Purchase-order and BACS billing is available on contact-us plans.

What counts as a seat?

Nothing. We bill on units and plan tier, not per user. Add as many people to your dashboard as you need.

What is a unit?

One unit is one Claude Sonnet input token. Other tokens convert to it at the rate the model actually costs: an output token counts as 5, an Opus token as roughly 1.7, a cached read as 0.1.

We do this because a token is not a token. Anthropic charges output at five times input, so a plain token allowance quietly means one thing for a summarising workload and something five times more expensive for a drafting one. Rather than write an allowance we could not honour for output-heavy work, we bill the thing that actually varies. Your dashboard shows units used, and every request reports the units it consumed.

Start on a trial, not a sales call.

Create an account, generate a key and send a request. Talk to us when you actually need to.

Start a free trial Read the docs

14 days free · cancel any time · nothing charged today

Questions about a specific compliance requirement? olly@dijitul.uk