Codex vs Cursor: which one's apps make money

Cursor wins on volume: ProvenStartups indexes 44 businesses whose founders name Cursor in the build story, against 10 that name Codex. Read that gap as adoption age, not a quality verdict — we have 44 Cursor cases and 10 Codex cases, and the Codex side is thin. Ten rows settle nothing, but they do show what people charge for and how well each claim holds up.

Contents

• ·Which tool's builders have the bigger proven numbers?
• ·What do the top Codex and Cursor cases actually earn?
• ·Where do the two cohorts cluster?
• ·What do the evidence grades say about each side?
• ·Which one should you pick?
• ·What can't these numbers tell you?

On that second point: 3 of the 10 Codex cases carry third-party-verified revenue, against 10 of the 44 Cursor cases — proportionally better, on a sample where one case swings the ratio ten points. Plan against the table, not the ratio.

Which tool's builders have the bigger proven numbers?

Cursor, and at the top it is not close. The Cursor cohort holds the largest verified figures here, including the tool's own maker, Anysphere, at $500M/yr with 60 people and a $9B valuation. The largest verified number on the Codex side is BridgeMind at $242,964 ARR.

Below Anysphere the Cursor side stacks up fast: Gravl at $440K/mo with 70K+ subscribers, Cursor Directory at $34,000–35,000/mo from a three-hour weekend project, SiteGPT at $13K MRR — all verified. The Codex side's headline claims are bigger than its verified ones: Super Demo at $250K+ MRR and Clarvo at $1M ARR are both founder-reported.

Our position: a verified $34K/mo beats an unverified $1M ARR as something to plan against. The grade is not an accusation; it states how much independent support sits behind a number.

What do the top Codex and Cursor cases actually earn?

Published figures run from $21K/mo to $500M/yr. Read the evidence column before the revenue column: ✅ third-party verified, 🗣 founder-reported, 📎 creator-relayed (a figure the video's creator sourced elsewhere), 🔮 unproven. Tier 1 means copy this now, Tier 2 replicable, Tier 3 watch the space.

Three rows sit in both cohorts, because founders name several tools inside one breakdown. Every case here also sits under our AI coding tools hub.

Where do the two cohorts cluster?

Cursor spreads across ten categories, Codex across four. The Cursor 44: 15 SaaS, 9 consumer apps, 5 scale references, 4 AI websites, 4 platform plugins, 3 directory sites, and one each of ecosystem tool, simple tool, AI service and cautionary tale. The Codex 10: 4 SaaS, 4 consumer apps, 1 plugin, 1 AI service.

The missing categories carry the signal. Codex has no directories, no ecosystem tools, no scale references. Cursor's list contains businesses built on top of the tool's ecosystem — Cursor Directory at $34,000–35,000/mo, Aura's template library at $15,000 MRR and 21,700+ users in about a month — plus the platforms next door, Windsurf at a $1.3B valuation and Replit at roughly $160M ARR.

Second-layer businesses are the clearest sign of a mature ecosystem, and only one side has them — a fact about calendars, not compilers. Solo operation looks identical: 28 of 44 Cursor cases and 7 of 10 Codex cases are one-person outfits.

What do the evidence grades say about each side?

Both cohorts run on founder claims. Cursor: 10 verified, 26 founder-reported, 4 creator-relayed, 4 unproven. Codex: 3 verified, 4 founder-reported, 2 creator-relayed, 1 unproven. The most common grade on either side is a number founders published about themselves that nobody independently checked.

So we weld the grade to the figure. The two biggest claims in the table — Revid at $600K+/mo, Super Demo at $250K+ MRR — are founder-reported and unsupported by anyone outside the business. The verified numbers are smaller and more useful: Prayer Lock reports $21K/mo from an app the founder says took three days to write; the work went into one video format that converted, and into onboarding. We rank tools this way in AI coding tools ranked by revenue.

Which one should you pick?

Pick on workflow, because this data does not separate the tools on outcomes: no business category appears on the Codex side that is missing from the Cursor side. Use the cases to choose a market instead. Three profiles, each tied to someone who already got paid.

1. 1.You have a manual process clients already pay for. Either tool. Profit AI is the template: a non-technical founder uploaded the spreadsheet he had kept for years, said build it, then read $147,000 of total revenue off the Shopify partner dashboard on camera.
2. 2.You want a consumer app in a niche that already has a competitor. Copy, then out-execute. The Codex ten hold two faith-based screen blockers — Prayer Lock at $21K/mo ✅ and PrayLock at $40K/mo 📎 — while Cursor's side has Gravl at $440K/mo ✅, which copied the market leader's interface and replaced its training engine.
3. 3.You want to sell to builders. Cursor, by default: every ecosystem case with money attached is on that side. We hold no Codex equivalent yet — either the opening or the warning, depending on how early you like to be.

What can't these numbers tell you?

Who failed. We only index businesses whose founders published a figure, usually on camera and usually while things were going well. The Cursor cohort holds exactly one cautionary tale; the Codex cohort holds none. That is a sampling artifact, not a safety rating.

• ·Attribution is fuzzy. Founders name several tools in one interview; three cases here belong to both cohorts.
• ·Sample size is doing real work. One new verified Codex case moves its ratios more than ten new Cursor cases move theirs. We nearly called the Codex side better documented, then counted the rows.
• ·Every figure is a snapshot. Topical Map AI went $0/mo → $5K → $7K → $10K/mo, made about $120,000 gross over 18 months, and sold for five figures. Growth curves end.

For the wider set, see 46 apps built with Cursor and what Cursor AI actually is.

Frequently asked questions

Is Codex or Cursor better for making money?

Neither, on this evidence. We index 44 revenue cases naming Cursor and 10 naming Codex, and every business category on the Codex side also appears on the Cursor side. The volume gap tracks how long each tool has been in wide use, not how much its users earn.

How many Codex and Cursor cases have verified revenue?

Three of the 10 Codex cases and 10 of the 44 Cursor cases carry third-party-verified figures. The rest are founder-reported, creator-relayed or unproven. Founder-reported is the most common grade on both sides, so most of what circulates about either tool is a self-published number nobody has checked.

Why do some projects appear under both tools?

Because founders name more than one tool when they describe a build, so three cases here sit in both cohorts. Profit AI is the clearest example: its build story credits Cursor by name and it still turns up in the Codex list. Treat tool attribution as a label on a story.

What is the largest verified number on each side?

On the Cursor side, Anysphere, the company that makes Cursor, at $500M a year with a $9 billion valuation; among products merely built with it, Gravl at $440K a month. On the Codex side, BridgeMind at $242,964 ARR. All three are third-party verified, which is why we lead with them.

Should I switch tools if my app isn't making money?

No. Nothing in either cohort suggests the editor is the constraint. Prayer Lock's app took three days to write, and the founder puts the work into a video format that converted and into onboarding. When revenue is flat, the honest diagnosis is usually distribution or positioning.

More in AI coding tools.