Cursor Changed Its Price Six Times in 12 Days

Last week I wrote that Cursor changed its price four times in one week. That post is now out of date. The count is six, across 12 days, and the pattern has stopped looking like an accident.

Same method as always: every price below is a diff between two daily snapshots of Cursor's public pricing page. Nothing here came from an announcement, because none of it was announced.

How many times has Cursor changed its pricing?

Six times between July 14 and July 25 — three complete round trips between $20 and $16 for Pro, and $40 and $32 per user for Team:

• Tue Jul 14 — Pro $20 → $16/mo, Team $40 → $32/user/mo. A clean 20% cut.
• Fri Jul 17 — Back to $20/$40, with the tiers renamed around personas and a new "Contact sales" enterprise tier.
• Sat Jul 18 — Down again to $16/$32.
• Sun Jul 19 — Back up to $20/$40.
• Thu Jul 23 — Down again to $16/$32, the same day Cursor Router shipped.
• Sat Jul 25 — Back up to $20/$40, with Pro repositioned around "Composer" access instead of "limited tab completions."

One reversal is a mistake. Three identical round trips in 12 days is a system. Cursor is running live price tests on its public pricing page, and the $16 figure keeps coming back because someone keeps putting it there.

So what: if you sell anywhere near Cursor, a single observation of their price is worthless for competitive positioning. The number you screenshot on a Tuesday may be gone by Thursday. What matters is the range they're testing — $16 to $20 — because that is the band they believe the market will bear.

Why is Cursor testing its price right now?

Because the thing it sells is changing, and this is the more interesting story. Three changes captured in the same 12-day window explain the price instability better than the price changes do:

• Cursor Router shipped July 23. It automatically picks between frontier and cost-efficient models based on task complexity, with Intelligence, Balance and Cost modes plus admin controls on Teams and Enterprise. It lets Cursor claim "best model" and "cost-controlled" at the same time — which, until now, were opposing promises.
• Agent limits became the tier ladder. The pricing page now differentiates plans by multiples of agent usage — 3x, 20x and 5x limits on Agent across tiers — alongside priority feature access. Not seats. Not completions. Agent throughput.
• Cursor Start launched for India at ₹649/month, tax inclusive, billed in INR with UPI or card. A locally priced entry tier is a different pricing motion entirely: geographic segmentation, not a discount.

Put together, Cursor is moving its unit of value from "an editor seat with autocomplete" to "how much agent work you get." That reset is why the headline price won't sit still: when you don't yet know what your unit is worth, you test the number in public until you do.

So what: the durable signal isn't $16 or $20 — it's that agent throughput is becoming the thing customers pay for in AI coding tools. If you price per seat in this category, that assumption has a shelf life.

What else moved across the stack

The same repricing-around-AI pattern showed up everywhere I watched this week.

Lovable ran what reads as an enterprise-readiness sprint: Gemini 3.6 Flash became the default model for apps' built-in AI features, plus Notion, Apify and Tally connectors, payments infrastructure work, external invites on Business and Enterprise, and build-secret visibility controls for editors. Connectors bring the integrations enterprise buyers ask for; access controls and secrets handling answer the security review. That combination is what selling upmarket looks like from the outside.

PostHog made two moves that look contradictory and aren't. Log ingestion got cheaper per unit — $0.25 down to $0.05/GB — while the free allowance dropped from 50 GB to 10 GB, an 80% cut, with extended retention options added on top. Cheaper at volume, tighter at zero. It also clarified per-event pricing at $0.00006 with 100k events free. Read together: make paid consumption easy to model and make the free tier stop absorbing serious workloads.

beehiiv removed token limits from Copilot across every pricing tier, deleting the hourly and weekly caps it had shipped weeks earlier — the full "we capped it, users hated it, we uncapped it" arc, visible only in the diffs.

Loops rewrote its pricing page to lead with "no separate charge for sending" and published explicit rate limits: 10 emails/second on free, 1,000/second on paid. Most email platforms bury throttling in docs. Putting it on the pricing page as a feature is deliberate positioning against the ones that don't.

The pattern worth naming

Four categories — AI coding, app builders, analytics, email — and all of them spent this week changing how they charge rather than what they ship. Usage replaces seats, AI capacity becomes the metered unit, and free tiers tighten while paid consumption gets cheaper per unit and easier to forecast.

If you're planning a pricing change this quarter, you're not doing it alone. Your buyers are being trained right now to expect metered, AI-inclusive plans — and to expect the number to move.

How this was made

Every fact above is a diff between two daily snapshots of a public page: pricing, changelog or blog. No press releases, no tips. Cursor never announced that it dropped its price to $16, three separate times. It just changed the page, and changed it back, and the only way to know is to have been looking every day.

That's the whole product. If checking competitor pages by hand isn't happening at your company — and it usually isn't — here are the practical ways to track competitor pricing changes, including the automated one.

ScrapRival watches your competitors' pages and sends one briefing every Monday. You can read a real sample digest without signing up, or start a 14-day free trial pointed at your own competitive set.

Next Monday: same treatment, and we'll see whether $16 comes back a fourth time.