Claude Code for UK solopreneurs: the daily workflow that actually ships

Claude Code is the single biggest productivity shift for UK solopreneurs in 2026. But it also makes it very easy to spend eight hours a day pair-programming with an AI and get nothing shipped.

This is the workflow I'd hand to a UK solopreneur starting Monday morning with Claude Code installed and a SaaS to build. Morning, afternoon, evening. Real cadence, UK time zones, GBP costs, no hype.

The core trade-off

You can treat Claude Code as a faster typist or as a junior engineer. The first mindset wastes 80% of the value. The second one ships.

A junior engineer needs:

• A clear goal written down before they start
• A plan before they touch code
• Context on the codebase
• Checkpoints where you review
• A definition of done

Your CLAUDE.md file and your plan-mode workflow are the two places you make that real. Everything else in this post is scaffolding around those two things.

The day at a glance

Block	Time	Output
Morning intake	08:00-09:00	Today's one ship target + plan
Deep build	09:00-12:30	Actual feature(s) in plan mode + execution
Lunch + review	12:30-14:00	PR review, cup of tea, walk
Ops + admin	14:00-15:30	Ticket triage, customer email, invoicing
Shallow build	15:30-17:30	Small features, tests, polish
Ship + log	17:30-18:00	Deploy, devlog, close the laptop

Three hours of deep build and two hours of shallow build. One hour of admin. The rest is thinking, reviewing, or living. If you try to do six hours of deep build five days a week, you will burn out by month three.

08:00 — Morning intake (one hour)

Walk, shower, coffee, whatever. Then 30 minutes at the desk doing three things:

1. Pick today's one ship target. One. Not three. One feature, bug, page or integration that will go from "not done" to "live on production" by 18:00. Write it at the top of your devlog in plain English.

2. Read the CLAUDE.md file. Not all of it — the sections relevant to what you're shipping today. Add anything you learned yesterday that's missing. Keep it under 150 instructions total. Long CLAUDE.md files degrade behaviour.

3. Check inbox and Slack for five minutes max. Anything that blocks today, deal with now. Anything else, note and move on. No rabbit holes.

UK solopreneurs have a specific morning advantage: the US hasn't woken up yet. You get 09:00-14:00 with zero inbound. Guard it.

09:00 — Deep build (three and a half hours)

This is the block that moves the product forward. It has a specific structure.

Step 1: Plan mode, not code

Open Claude Code. Press Shift-Tab twice to enter plan mode. Describe the feature like you're briefing a senior engineer who has never seen the codebase.

Example prompt:

"Plan an implementation for user-subscription cancellation. Requirements: UK GDPR-compliant, must keep data for 30 days post-cancel for restoration, cancel immediately (no grace period), Stripe subscription cancellation via webhook, confirmation email via Resend. The codebase is Next.js 16 App Router + Supabase + Stripe, server actions already exist for subscription creation in app/actions/subscription.ts. Don't write code yet — plan the steps and flag any ambiguity."

Claude returns a plan. Iterate on it for 10-30 minutes. Push back where the plan skips steps or assumes something. Add acceptance criteria in writing.

Step 2: Execute in small steps

Exit plan mode. Ask Claude to implement step 1. Review the diff. Ask for step 2. Review the diff.

Do not paste "implement the whole thing" and walk away. You will get something that compiles and doesn't work. Review each diff the same way you'd review a junior engineer's PR.

Rules of thumb:

• One PR per feature, not one per day.
• Tests before merge for any business logic — auth, payments, subscription state, data access.
• Claude writes the tests too, but you review them. Ask: "does this test actually fail if the implementation is wrong?"
• Commit after every green checkpoint. If Claude breaks the next step, git reset beats an hour of untangling.

Step 3: Exit on time

12:30 is a hard stop. Ship what you have to a branch. If the feature is 70% done, commit it, push, and come back after lunch with a fresh brain. Forcing the last 30% with a tired brain is where bugs go to ship.

12:30 — Lunch and review (90 minutes)

Two halves.

Lunch — leave the desk. Not with a laptop, not with a phone scrolling Twitter. 30-45 minutes of actual rest. UK weather permits a walk for roughly 40% of the year; take it when it does.

Review — 30 minutes on the laptop at a different desk if possible. Read the code you wrote this morning. Not to fix it — to notice things. What's ugly? What's confusing? What did you learn about the codebase that should go in CLAUDE.md?

Write one line in a notes file for each: codebase insight, pattern I want to repeat, pattern I want to avoid. Feed the last two into CLAUDE.md during tomorrow's morning intake.

14:00 — Ops and admin (90 minutes)

This is the UK solopreneur tax. You don't have a co-founder doing the boring stuff, so it goes here.

Run through a fixed list. Don't let it expand.

• Customer support — reply to every email from the last 24 hours. Use a templated opener but write every answer personally.
• Payments and invoices — Stripe dashboard check, any failed payments, any refund requests, monthly invoicing if due.
• Admin inbox — HMRC, Companies House reminders, bank, domain renewals, any agency or tool renewals. Bin anything not time-sensitive.
• Accounting — 15 minutes in FreeAgent / Xero, categorise the week's transactions. Once a week becomes once a quarter and then a nightmare.
• Content commitments — newsletter, blog post, social. If today's the day, write for 30 minutes. If not, skip.

Set a 90-minute timer. When it ends, stop, even if the list isn't clear. Overflow goes to tomorrow.

15:30 — Shallow build (two hours)

Afternoon brain is worse at deep problems. It's better at:

• Small features (a settings toggle, a copy change, a new field)
• Tests for code you wrote this morning
• Polish — empty states, loading states, 404 pages, error messages
• Refactors with clear before/after
• Small bug fixes with a clear reproduction
• Writing or updating your CLAUDE.md

Claude Code is excellent for these. You can run two or three of them in the same session without losing the plot. The deep-build discipline (plan mode, review each diff) still applies, but at lower intensity.

17:30 — Ship and log (30 minutes)

Three things.

1. Ship. Deploy to production. Smoke-test the feature in prod. Fetch the live URL with curl. If it's broken, fix it or revert it. Do not leave a broken production overnight — you'll wake up to a customer email and cortisol before coffee.

2. Devlog. Write three lines in a daily log: what shipped, what didn't, what's on tomorrow's list. Five minutes, not fifty. Future-you in six months will thank present-you.

3. Close the laptop. UK solopreneurs without this rule burn out in year one. Set a hard 18:00 stop. If you want to come back at 21:00 for one hour, that's fine — but it's an hour, not "I'll just finish this."

Real UK costs for this stack

Monthly spend for one solopreneur running this workflow:

Tool	Monthly cost	Notes
Claude Code (Pro or Max)	GBP 17-85	Max tier covers typical solopreneur usage
Cursor (optional second tool)	GBP 16	If you like a GUI editor alongside Claude Code
Supabase Pro	GBP 20	Free tier works for first 6-12 months
Vercel Pro	GBP 16	Free tier works for validation, Pro when you have customers
Stripe	Variable	1.5% + 20p per UK card transaction
Domain	GBP 12/yr	.co.uk or .com
ICO registration	GBP 40-60/yr	Mandatory for most UK SaaS
Accounting (FreeAgent)	GBP 19	Free with some business bank accounts
Email (Resend)	GBP 16	Free for first 100 emails/day

Total: GBP 105-175/mo before revenue. Most UK solopreneurs get their first paying customer within 60 days on this stack.

The CLAUDE.md discipline

Your CLAUDE.md file is the single highest-leverage artefact you own. It persists across sessions, informs every interaction, and compounds over time.

Keep:

• Under 150 instructions. Long files degrade model adherence.
• Code conventions (naming, file structure, imports).
• Stack specifics (Next.js 16 App Router, React Server Components, Server Actions).
• Do/Don't rules that come from mistakes you've made (e.g., "Never use useEffect for data fetching — use Server Components or use hook").
• UK-specific rules ("All currency shown as GBP with £ prefix", "VAT-inclusive pricing shown for consumer tiers").

Prune:

• Anything out of date
• Anything that's now in the codebase as a convention (e.g. ESLint rules — Claude reads the config)
• Anything you haven't referenced in the last month

Review monthly, on the last working Friday. 20 minutes.

Common failure modes

Failure mode 1: Plan-mode skipping. You jump straight to "write me a subscription cancellation flow" and get a generic implementation that half-fits your codebase. Fix: always open in plan mode first for anything longer than a 10-line change.

Failure mode 2: Not reviewing diffs. You accept everything Claude writes because it compiles. By week three, the codebase has three competing patterns for the same thing. Fix: review every diff like a PR, even for one-line changes.

Failure mode 3: Tests that don't fail. Claude writes a test that passes whether the implementation is right or wrong. Fix: deliberately break the implementation after writing a test and confirm the test fails. This takes 10 seconds and catches 90% of useless tests.

Failure mode 4: Endless shallow-build afternoon. You spend four hours on polish and skip the admin block. Fix: the 90-minute admin timer is non-negotiable. Shallow build only after admin is cleared.

Failure mode 5: The "just one more hour" trap. You extend deep build into shallow build, shallow build into admin, admin into evening. You've been at the desk for 10 hours and shipped nothing. Fix: the time blocks are constraints, not suggestions.

Where this workflow breaks

Breaking: customer-facing emergencies. A production outage or an angry customer jumps the queue. Deal with it in the deep-build block if it's during work hours. Re-plan the day afterwards, don't try to compress everything.

Breaking: co-founder or contractor sync. If you're not actually solo, you need a sync block at the start of the afternoon, not in the morning. 30 minutes at 14:00 works for UK/UK teams; 16:00 for UK/US.

Breaking: consulting on the side. If you're funding your build with client work, split the week: 3 days solopreneur, 2 days client, instead of splitting the day. Context-switching kills deep build.

A note on Claude Code vs Cursor vs other tools

This workflow works with any capable AI coding tool. Claude Code is what I'd recommend in 2026 for UK solopreneurs because:

• Plan mode is the single best feature for solopreneur discipline. It forces the "senior engineer briefing" habit.
• CLAUDE.md is first-class. Cursor's equivalent exists but is less mature.
• Terminal-first fits a keyboard-driven workflow. You can still use any editor alongside.
• GBP pricing is transparent — Anthropic publishes UK pricing, no dollar-conversion games.

You could run this exact day with Cursor's agent mode or Bolt in some configurations. The workflow matters more than the tool. The tool makes the workflow cheaper.

Key takeaways

• One ship target per day beats a long to-do list; write it at 08:00, ship it by 18:00.
• Plan mode first, execute second; treat Claude Code like a junior engineer, not a faster typist.
• Three hours deep build, two hours shallow build, 90 minutes admin - this is the ceiling for one solopreneur, not the floor.
• CLAUDE.md under 150 instructions is the highest-leverage artefact you own; review monthly.
• UK solopreneur monthly stack costs GBP 105-175/mo before revenue; most get a first paying customer inside 60 days.

FAQs

Is Claude Code Pro enough for a UK solopreneur? For most solopreneurs, yes, until you hit usage ceilings. Max tier (GBP 85/mo) covers heavy daily use. Start on Pro, upgrade when you see rate limits.

Do I need Cursor if I use Claude Code? No. Many UK solopreneurs run Claude Code alone with any editor (VS Code, Zed, Neovim). Add Cursor only if you prefer a GUI agent mode for specific tasks.

How big should CLAUDE.md be? Under 150 instructions. Research shows LLMs follow fewer instructions more reliably than many poorly.

What if I can't do three hours of deep build? Most people can't in the first month. Start with 90 minutes, grow it weekly. The block is an aspiration for the first quarter.

How do I handle HMRC and tax time with this workflow? The 90-minute afternoon admin block includes 15 minutes a week on accounting. If you do that, your quarterly VAT return (if applicable) and annual self-assessment are a one-day exercise, not a two-week nightmare.

[!info] Next step Want a fully researched UK business opportunity to build with this workflow? Read this week's free report for keyword data, competitor analysis, and builder prompts.