Cursor Priced India at 649 Rupees: How to Choose and Defend a Second Country Price

If you are considering a cheaper price for a poorer country, start from a published number rather than a guess, then decide what you will remove from the product to defend it. The World Bank's purchasing power parity data gives you the first half in about ten minutes. Cursor launched an India-only plan today at 649 rupees a month against a $20 global plan, which is a steeper cut than most companies dare and a shallower one than the PPP ratio implies. Below: the ratios for ten countries, what Cursor's number works out to, the two mechanisms that actually ship a second price, and the hole each one leaves for arbitrage.

What Cursor shipped

Cursor Start is 649 rupees a month, tax inclusive, India only, paid by UPI or card, on monthly auto-renewal, per the company's changelog (opened 28 July 2026). It includes access to Grok 4.5 and Composer, always-on cloud agents, the iOS app, plugins, MCP servers, hooks and skills. Free users in India upgrade from the dashboard. The global pricing page (opened 28 July 2026) does not list Start at all: it shows Hobby free, Individual Pro at $20 a month with frontier model access, Pro+ at three times Pro's agent limits, Ultra at twenty times, and Teams at $40 per user per month.

Two structural choices are visible there. The plan is invisible outside India, and it is a different product, not the same product at a lower price. Both matter, and we will come back to why.

A defensible starting number

The World Bank publishes a PPP conversion factor for GDP, in local currency units per international dollar, alongside the official market exchange rate in local currency units per US dollar. Divide the first by the second and you get a price level ratio: roughly, what a dollar's worth of output costs in that country relative to the United States. The table below is our computation from both 2025 series, retrieved from the World Bank API on 28 July 2026. The last column applies the ratio to a $20 anchor price.

Methodology and its limits, because both belong in the same paragraph. The inputs are the World Bank indicators PA.NUS.PPP and PA.NUS.FCRF, latest available year 2025 for every country listed; the ratio and the implied price are ours. This measures the price level of a whole economy, not what a developer at an outsourcing firm in Bengaluru will pay for a coding tool, and those two things diverge sharply for products bought by people whose salaries are set by international demand. Treat the column as a floor for the discussion, not as a price.

Where Cursor landed against it

Our arithmetic, at the World Bank's 2025 average rate of 87.158 rupees per dollar: 649 rupees is $7.45 gross of tax, which is 37% of the $20 Pro price. Cursor states the price is tax inclusive, so the amount it keeps is lower still; at India's 18% goods and services tax on this category the ex-tax figure would be around 550 rupees, or about $6.31, which is 32% of Pro. We have not verified that tax rate against a primary source this run, so treat the ex-tax number as an estimate and the $7.45 as the solid one.

Either way, the PPP ratio for India would put the price at $4.60. Cursor stopped roughly a third of the way short of that, and closed the remaining gap on the product side instead: Start is built around Cursor's own models rather than the frontier models on Pro. That is the trade worth copying. A price cut and a feature cut do different jobs. The price cut buys reach; the feature cut is what stops the cheap plan from cannibalising the expensive one and gives you an answer when a customer in Ohio asks why they pay three times what someone else pays for the same thing.

Currency localisation is not price localisation

These get conflated constantly, and they are separate products with separate mechanics.

Stripe Adaptive Pricing (opened 28 July 2026) converts your existing price into the customer's local currency across 150+ countries, holding an exchange rate for 24 hours. You pay nothing for it; the customer pays a 2% to 4% conversion fee built into the displayed price. It works with subscriptions for card, Link, Apple Pay and Google Pay across borders, and you switch it on at dashboard.stripe.com/settings/adaptive-pricing. Read the exclusions before you plan around it: it is not available for Elements with the Payment Intents API, not for manual capture or custom amounts, not for prices that already declare currency_options, and not for businesses based in India. Refunds go back in the customer's original payment currency.

That is conversion. It removes the friction of a foreign currency and it does not make anything cheaper.

Paddle covers the other half. Its localized pricing documentation (opened 28 July 2026) describes unit_price_overrides on a price entity, where each override carries an array of country_codes and a unit_price with its own amount and currency. Precedence runs override first, then automatic conversion, then the base price. That is a genuine second price, set by you, per country.

The arbitrage hole, in the vendor's own words

Paddle documents exactly how a buyer's country is decided: "When opening a checkout, Paddle uses geolocation to estimate where a customer is buying from." And then, one line later: "If a customer changes the preselected country, Paddle gets localized prices for the country they selected."

That is the whole exploit. No VPN required for the version that matters; the country selector is part of the checkout. Which means an override is a posted discount available to anyone willing to click a dropdown, unless you add a second gate. The gates that actually work are the ones tied to something harder to fake than an IP address:

• A local payment method. Cursor taking UPI for its India plan is not only a convenience choice. A payment rail that requires a local bank account is a stronger residency signal than geolocation.
• A tax identifier. If your regional plan requires a local business tax ID that you validate, casual arbitrage stops being casual.
• A different product. The most robust gate of all, because it does not need enforcing. Nobody in London switches to the India plan to lose frontier model access.
• No annual prepay at the regional price. Cursor's Start plan is monthly. A month of mispriced subscription is a rounding error; a prepaid year is a real loss.

One legal note before you ship anything clever. Pricing by country from a published, uniform list is ordinary commerce. Pricing an individual from data you hold about them is a different thing, and three US states have now written that distinction into statute, which we covered in the piece on surveillance pricing. If your regional price is decided by an account attribute rather than by a country list, read that first.

Here is the rule we would apply. Do not build a second price until one country is already producing enough signups on the free tier to be worth measuring, and until you know which feature you are willing to remove. Absent both, you are not running a pricing experiment, you are running a discount, and a discount with a country name on it is the kind of decision that is easy to announce and expensive to reverse. The infrastructure question, which processor and what it costs you, is downstream of that and covered in our comparison of what platforms charge for using your own gateway.