50 Apps Built With Claude Code That Have Revenue Evidence
ProvenStartups tracks 50 apps and projects built with Claude Code that are tied to real revenue evidence, and 41 are solo-run. Twelve publish a clean…
ProvenStartups tracks 50 apps and projects built with Claude Code that are tied to real revenue evidence, and 41 are solo-run. Twelve publish a clean monthly figure: the median across the full matching set is $20K/mo, with a $2K/mo to $500K/mo range. The useful answer is not “ship more apps”; it is to choose a narrow paid problem and inspect the evidence class before copying anything.
Contents
What the 50-project cohort actually shows
This cohort is mostly solo, commercially varied, and far less uniform than “Claude Code SaaS” implies. It contains 50 projects, 41 run by one person, but only 12 disclose a clean monthly result. Those 12, not a hand-picked sample, produce the $20K/mo median and the full $2K/mo to $500K/mo range.
SaaS is the largest category with 14 cases, followed by nine AI services and eight consumer apps. The remainder includes five directory sites, three platform plugins, three AI websites, two simple tools, two cautionary tales, and one each in scale reference, AI content, ecosystem tool, and digital publishing.
There is also a grading caveat worth stating plainly. The supplied cohort split reports nine third-party-verified cases [V], zero founder-reported [F], zero creator-relayed [C], and zero unverified [U]. It does not disclose why the remaining records are absent from that aggregate split, so ProvenStartups does not invent an explanation.
That restraint matters. Across the full index of 406 startup ideas, ProvenStartups separates 57 [V], 184 [F], 121 [C], and 44 [U] records. A revenue claim and a verified revenue claim are different inputs, even when the product looks equally polished.
Revenue comparison: six Claude Code projects
The strongest reusable pattern is a narrow workflow with an obvious buyer, not a particular framework or launch channel. The table mixes verified outcomes, founder reports, and an undisclosed result deliberately. Read the grade beside every figure first; then compare category and difficulty. These are examples from the allowed case set, not the entire 50-project cohort.
| Project | Reported result | Category | Difficulty |
|---|---|---|---|
| AEO Service | $2,000/mo retainer from one client [F] | SaaS | 1/5 |
| Minea / DropMagic | Minea peaked at $750K MRR; DropMagic reached $45K MRR in four months [F] | SaaS | 4/5 |
| Cursor | $500M/yr, 60 people, $9B valuation [V] | Scale reference | 5/5 |
| AI App Factory | Revenue undisclosed; one-time purchases start at $0.99 [U] | Consumer app | 2/5 |
| Payout | $20K/mo, reached in 50 days [V] | Consumer app | 3/5 |
| Subscribr | $30K/mo [V] | SaaS | 3/5 |
Do not average this table. Cursor is a scale reference, the app factory discloses no revenue, and Minea’s peak is not a current run rate. Payout at $20K/mo [V] and Subscribr at $30K/mo [V] are cleaner operating comparisons because both publish monthly figures with third-party verification.
The low end is useful too. One paid AEO client produced a $2,000/mo retainer [F]. That is weaker evidence than [V], but it describes a smaller, testable sales motion rather than asking a solo developer to imitate a company doing $500M/yr [V].
What we would build, refuse, and verify
We would build a narrow tool or productized service with one measurable job, one reachable buyer, and a manual fallback. We would refuse to start with a mass-app portfolio, a generic directory, or an “uncapped” venture studio thesis. Those models can work, but the cited records do not provide equally strong, current revenue proof.
Three routes deserve different treatment:
- 1.Start service-first. The Claude Code local SEO service reports $5,000+ cumulative digital-product revenue plus client retainers described as several thousand dollars monthly [F]. The figure is founder-reported, but the offer can be sold before automation is complete.
- 2.Turn a specific workflow into software. Payout reached $20K/mo [V], while Subscribr reports $30K/mo [V]. Both have a defined user action and a result customers can understand without learning a new category.
- 3.Treat distribution-heavy models as hypotheses. The Claude Code and Crawl4AI directory targets $2K–$10K/mo [C]; that is a target, not earned revenue. The AI Venture Studio describes profitability as theoretically uncapped [C], which is not a usable forecast.
Before building, read Anthropic’s official Claude Code documentation for the product surface and Anthropic’s engineering write-up on Claude Code practices for operating patterns. Then inspect ProvenStartups’ grading method. Tool competence reduces implementation friction; it does not validate demand.
Our practical filter is simple:
- ·Prefer a paid pain over an interesting demo.
- ·Require a buyer list before a feature list.
- ·Separate earned revenue from targets, peaks, valuations, and cumulative sales.
- ·Match scope to the operator. A 1/5 service and a 5/5 scale reference are not adjacent starting points.
Where the data contradicts the popular Claude Code playbook
The popular claim is that Claude Code changes the startup game by making it rational to generate many apps and wait for one to hit. This dataset says the defensible edge is narrower: most matching projects are solo-run, yet fewer than one quarter publish a clean monthly figure. Shipping volume is visible; durable revenue proof is scarce.
The clearest warning is the AI App Factory. It says 90% of paying users are overseas and purchases begin at $0.99 [U], but it discloses no revenue total. That may be a valid portfolio strategy. It is not evidence that mass production beats one focused product.
The site-wide data points the same way. Among 106 cases with a clean monthly figure, 54 sit between $10K and $100K monthly, while 26 exceed $100K monthly. Yet ProvenStartups also files 38 documented failures as cautionary tales. Revenue evidence should narrow a decision, not erase survivorship bias.
Claude Code is prominent, but it is not uniquely predictive. Across 211 distinct projects mentioning at least one coding tool, Claude Code appears in 50 cases, compared with 100 for ChatGPT, 46 for Cursor, 40 for Bolt, and 19 for Lovable. Those are case counts, not a controlled productivity benchmark.
FAQ
The short version: Claude Code has credible revenue cases, but the dataset does not justify treating the tool as the business model. Use the cohort to select a problem shape, revenue model, and evidence threshold. Do not copy an outlier’s headline figure without checking whether it is monthly, current, verified, and achievable by a solo operator.
Are all apps built with Claude Code SaaS products?
No. Only 14 of the 50 matching projects are categorized as SaaS. The cohort also includes AI services, consumer apps, directories, plugins, websites, simple tools, publishing products, and two cautionary tales. That spread is useful because Claude Code is an implementation tool; it does not force a subscription model or a browser-based product.
How much do Claude Code projects make?
Only 12 projects in the full matching cohort publish a clean monthly number. Their median is $20K/mo and their range is $2K/mo to $500K/mo. Do not apply that median to all 50 projects: most do not provide a clean monthly figure, and the disclosed set can overrepresent businesses willing to publish results.
Which case is the best solo-founder benchmark?
Payout is the cleanest compact benchmark here: $20K/mo reached in 50 days [V], a 3/5 difficulty consumer app, and a specific paid outcome. AEO Service is a smaller validation model at $2,000/mo from one client [F]. One is stronger evidence; the other may be a cheaper test.
Does a [V] grade mean the business is safe to copy?
No. [V] means the stated figure has third-party verification; it does not guarantee retention, margins, market size, or future performance. [F] is founder-reported, [C] is relayed by a creator, and [U] is unverified. The grade tells you how much confidence to place in the claim, not whether the idea fits your distribution.
Is Claude Code the best AI coding tool for startups?
This dataset cannot answer that. It records 50 Claude Code cases and 46 Cursor cases, while ChatGPT appears in 100, but those counts are not experiments with controlled teams and tasks. Use the tool that lets you review, test, and ship maintainable code; choose the business using customer access and graded revenue evidence.