Bespoke software and AI agents
Senior engineering, multiplied by agents.
Software and AI agents for founders and growing businesses, built in weeks. Agents do the heavy lifting and senior engineers lead every build, but you stay in control: you own everything, agents never touch your live systems, and nothing ships without a person’s sign-off.
- Code and data in your own accounts
- Agents cannot reach production
- A person signs off every release
Free scoping. A first view within two working days. See what we offer
What clients come to us with
Built around how you work, shipped in weeks.
- 01
Off-the-shelf never quite fits
Generic tools bend your process to the software, and spreadsheets fill the gaps.
We build around how you work, and you own the result.
- 02
Agencies are slow and expensive
A long timeline, a large team to pay for, and a handover that leaves you dependent.
Senior engineering with agents, at a fixed price.
- 03
AI on its own is a liability
Unsupervised agents produce work nobody can stand behind.
Every agent works inside written controls, and every action is recorded.
What we offer
Four services, each with a clear deliverable.
Bespoke software builds
A platform or product built from your brief, on a proven, modern stack, delivered to production.
You get
- Working software in your own accounts
- Tests, documentation and handover
- A person approves every release
- Length
- 1 to 8 weeks
- Best for
- A first product or one feature
Custom agents
An agent built around one defined task, answering from your documents and working in your tools.
You get
- Connected to your systems and files
- Every action logged with its approver
- Clear rules for when it asks a person
- Length
- 3 to 6 weeks
- Best for
- One repeated, rule-bound task
Operating platforms
One internal system for clients, pipeline, finance and documents, with agents that work across them.
You get
- A back office built around your work
- The best AI model for each job
- Dashboards on the numbers that matter
- Length
- 6 to 8 weeks
- Best for
- Teams outgrowing spreadsheets
Security and hardening
Review, test and harden your software, including products built fast with AI tools.
You get
- A signed scope before testing starts
- Every finding reproduced and ranked
- Fixes, tests and a retest
- Length
- 1 to 3 weeks
- Best for
- Before a launch, a raise or a sale
Built it fast with AI tools?
Make it safe before a launch, a raise or a sale.
Apps built with Lovable, Bolt, Cursor, Replit or v0 work on the happy path. We find what they miss, fix it, and prove each fix with a test.
1 to 3 weeks. Quoted to scope.
Ways to work together
Start small. Scale when it works.
Four services, and several ways to engage. Scoping a build you already have in mind is free, and everything else is priced in writing before any paid work begins.
- 01
Agent Readiness Review
Where agents can safely take on work across your business, with a costed roadmap.
- Length
- 2 weeks
- Price
- Fixed fee
- 02Most start here
Proof sprint
One bounded feature, workflow or working prototype, live in your accounts.
- Length
- 1 to 2 weeks
- Price
- From £3,000
- 03
Platform build
A full platform or agent system, delivered in phases, each priced before it starts.
- Length
- 6 to 8 weeks each
- Price
- Per phase
- 04
Run and improve
Hosting, monitoring, agent tuning and support with agreed response times.
- Length
- Monthly
- Price
- Quoted to scope
Founder partnership
A reduced fee plus a small equity stake, for early-stage companies.
Launch warranty
In every build: any defect against the agreed acceptance criteria, found in the 30 days after release, is fixed at no cost.
How we build
From a plain brief to production, in six steps.
Agents do the volume. We make the decisions, and you see working software every week.
Brief
You describe the outcome in plain language.
Who acts: You.Plan
Agents draft it. We check it and set the price.
Who acts: Agents, checked by our team.Approve
You approve the plan and the price before the build starts.
Who acts: You.Build
Agents build in parallel, each in its own separate workspace.
Who acts: Agents.Verify
Agents test and review. We read every result.
Who acts: Agents, checked by our team.Ship
A named engineer ships it. Agents have no access.
Who acts: Our team.
Plan and approve
A fixed price before any paid work
Scoping is free. You approve the plan, the price and the risks before the build starts.
First release
1 to 8 weeks
A single feature lands in 1 to 2 weeks. You see working software every week.
After launch
30 days of free defect fixes
Then a Run and improve plan, or a full handover to your team.
What we build
Software that runs real businesses.
Products
Consumer and member experiences.
Platforms
Accounts, payments and permissions.
Data systems
Sources of truth that explain every number.
AI and agents
Assistants that cite their sources, and ask first.
Internal tools
The software a company actually runs on.
Integrations
The services you already use, wired in properly.
We work with founders and growing businesses in any sector, and have shipped live systems in music, entertainment, sport, education, life sciences and healthcare. References on request.
Controls
Clear controls, at every level of autonomy.
What each agent may do is agreed with you in writing before it runs, and enforced by access controls. Merging, deploying and touching production are impossible for an agent by construction, not merely disabled.
Always needs a person
- Payments
- Deleting records
- Changes to customer records
- Releases to production
Never allowed
- Production access for the agents that write your code
- Working outside the systems you approve
- Searching sources you have not approved
Always recorded
- Which agent acted, and the sources it used
- What it did, and who approved it
- An answer its sources do not support goes to a person
Autonomy is earned with a track record.
Bounds start narrow. You decide when they widen.
- Observe
- Suggest
- Act with approval
- Act and report
- Act without sign-off
Payments, deletions, changes to customer records and releases always wait for a person, however good the track record. We run the security review we sell on our own builds. How we keep agents away from production.
Insights
Straight answers to the questions buyers ask.
Bespoke or off-the-shelf software: should you build or buy?
When to buy off-the-shelf software, when to build bespoke, and how to buy the core and build the edge, with the costs to compare over three years.
How do you check Supabase row-level security in an AI-built app?
How to check row-level security in a Supabase app built with AI tools: the queries to run, the mistakes to look for and how to test it as a real user.
How long does it take to build an MVP or a custom platform?
Realistic timelines for a feature, a first product and a custom platform, what makes projects slip, and how to ship sooner without cutting corners.
Questions
What clients ask us first.
References and a live demonstration on request. Every question we are asked.
Who owns the code?
You do, from the first commit. Your code lives in your repositories and your data in your cloud accounts from the first commit. There is nothing to hand back, because it was never ours, and the system stays with you if we ever part. There is no per-seat pricing.
How is it priced?
Scoping a build you have in mind is free. Every piece of paid work has a fixed price in writing before it starts, and a proof sprint starts from £3,000. Any later change is re-scoped and re-priced in writing.
How quickly will we hear back?
Within two working days you get a first view, and within a week a written plan, a fixed price and the risks. A single feature lands in 1 to 2 weeks, and you see working software every week of a build.
Do we need a technical brief?
No. Describe the outcome you want in plain language. Turning it into requirements, architecture and a plan is our job, and you approve the plan before anything is built.
What do you build with?
Next.js, React and TypeScript on Supabase or AWS for new products, and Laravel with PostgreSQL where you already run PHP. For AI we use Anthropic, OpenAI or open models, the best one for each job, and we check what each provider keeps before it touches your data.
What happens after launch?
Any defect against the agreed acceptance criteria, found in the 30 days after release, is fixed at no cost. Then you choose: a Run and improve plan with agreed response times, or a full handover with documentation your engineers can use.
Is the security review a penetration test?
No. It is an engineering-led review, with every finding reproduced, fixed and retested. If a customer or regulator needs a certified penetration test, we will say so and work alongside the firm you choose.
Start a conversation
Tell us what you need built.
Send a brief. It does not need to be technical. Within two working days you get a first view, and within a week a written plan, a fixed price and the risks.