Opportunities › Scaling app development beyond AI code generators › Proposed app
Keelwright
AI app building with cost ceilings, an architecture map, and code you own
- Score
- 46
- confidence 44
- Evidence
- 2
- items on the page
- Rivals
- 1
- named in sources
- First version
- 45 d
- estimate
- Type
- Desktop app
- PWA: runs in the browser, installs on Mac and Windows
- Market since
- 2023
- Emerging market · AI estimate
ScreensConcept
Desktop app: Developers work in sustained sessions reviewing architecture maps, setting cost ceilings, editing code, and coordinating with team members on projects. Sample screens on a computer, in the proposed brand. Their numbers are sample data.
Project budget
Under budget$18.40
About this appAI proposal
Keelwright is an AI app builder for web projects that keeps going once the app is no longer a prototype. You set a spending ceiling per project, see what each prompt costs before it runs, and the AI works inside that limit.
As the codebase grows, Keelwright keeps a map of its structure and flags the places where generated code is getting tangled. It suggests splits and cleanups before they turn into rewrites.
The code lives in your own Git repository in a standard Next.js and TypeScript layout. You can edit by hand at any moment and the AI picks up your changes. If you stop using Keelwright, you keep a normal project.
- Tagline
- Generate with AI. Know the cost. Keep the code.
- Voice
- PlainSteadyPractical
- Why this name
- A budget is a spending limit, and a keel keeps a boat steady as it grows. Keelwright suggests a builder who keeps your project upright as it scales.
In the first version
Members say which features they would use. It tells PWAS what people want; it changes no score.
Hard cost ceilings
Set a monthly or per-project limit. Each prompt shows an estimated cost first, and work stops or asks before crossing the ceiling.
No member yet
Architecture map
A live view of modules, routes and data models, with warnings where generated code is duplicated or tightly coupled.
No member yet
Scoped edits
Limit each AI request to chosen files or modules, so context stays small, changes stay reviewable, and cost stays down.
No member yet
Hand-edit friendly
Work in your own Git repo. Manual commits are read back in, so the AI never overwrites your changes without a diff.
No member yet
Brand identityAI proposal
Check the name against trademark registers before using it.
Palette
#14352EDeep Pine · primary#F6F3ECPaper · surface#1B1F1EInk · text#E8A33DAmber · accent#6B8F71Moss · data
Typography
Aa
Keelwright
ABCDEFGHIJKLMNOPQRSTUVWXYZ 0123456789
Generate with AI. Know the cost. Keep the code.
PricingAI proposal
The material gives no competitor prices. These tiers are a starting point to test, built on the usage-based and per-seat ideas in the opportunity.
Free
Free
- One project
- Architecture map
- Cost estimates before each prompt
- Small monthly AI allowance
Builder
$29 a month
- Five projects
- Hard cost ceilings
- Scoped edits and refactor PRs
- Git sync with hand-edit read-back
Team
$24 a month
- Per developer seat
- Shared project budgets
- Spend reports by person
- Unlimited projects
What it could be worthEstimate
What a software business making this much sells for today, from real deals. Pick a monthly revenue.
If it makes
Worth about
$194,000
Sale range $132,000 to $263,000
$120,000 a year in revenue
From 87 software businesses listed and sold on Empire Flippers in the last year, revenue vetted by the broker: they sell for about 19× their monthly revenue, most between 13× and 26×.
Measured on profit, the middle deal is 32× monthly net profit: about 2.7 years of profit.
An estimate, not an offer or financial advice. A real price depends on profit, growth, churn, how much of the work is the owner's, and the buyer.
How it would growAI proposal
How the product would reach new customers through its own use, and how the first ones would be found. A plan to test, not a result.
- Built-in loop
- No loop in the product. Reached through developer communities, content on scaling AI-built apps, and word of mouth. Output is a normal repo, so it carries no visible Keelwright branding by default.
- Do they become customers?
- Test whether developers who hit cost or complexity limits in other AI builders will pay to move an existing project in. Measure import-to-paid conversion.
- First customers
- Solo developers and teams of two to five who built a web app with an AI generator and are slowing down because of cost or code complexity.
- Where to reach them
- Hacker News and Show HN: Post a write-up on moving an AI-built app to maintainable code, with a free repo import and cost map.
- Indie developer communities: Offer migration help in forums and Discords where people discuss AI builders hitting limits.
- Founder-led outreach: Contact developers who publicly complain about AI builder costs and offer a free import audit.
- First milestone
- Ten teams import a real project and five pay for a paid plan within 60 days.
Why people already pay for this
- Public claimAI code generation tools become too expensive and complex to continue building with as applications scale beyond simple projects
- Public claimAI code generation tools are easy to get started with, suggesting they have found market fit for initial development
Tech stack
Every proposal builds on the same PWA base, installable from the browser on any phone or computer. Dashed boxes are what this app adds.
- AppNext.js and TypeScript
- PWAManifest and service worker
- UITailwind CSS
- Data and authSupabase Postgres
- PaymentsStripe
- EmailResend
- HostingVercel
- This app · AI model accessMulti-provider LLM gateway with token meteringNeeded to estimate and cap cost per request and to switch models by price.
- This app · Code analysisTypeScript AST and dependency graph parserBuilds the architecture map and detects duplication and circular imports.
RisksAI inference
- CommercialDevelopers may accept costs of existing builders or move to editor-based assistants, so the demand for a separate budgeted tool is unproven.
- AIReliable cost estimates before a prompt runs are hard, since model output length varies. Poor estimates would break the main promise.
- TechnicalReading hand edits back into AI context and importing repos from other tools with differing structures is complex.
- PlatformDepends on third-party model providers whose prices and terms can change, which may squeeze margins under fixed ceilings.
- DataHandling customers' source code requires strong access controls and clear storage and training policies to earn trust.
Everything marked AI proposal or AI inference was written by a model from the evidence on the opportunity page. The proposed name must be checked against trademark registers before it is used.