Cursor Start India pricing: ₹649/month, and why it works
Cursor's first India-only plan runs Composer 2.5 and Grok 4.5 for ₹649/month with UPI payments. What you get, what you lose, and why the price is sustainable, not a loss leader.
TL;DR: Cursor launched its first country-specific plan on July 28, 2026, days before SpaceX’s $60 billion acquisition closes. Cursor Start costs ₹649 per month for developers in India, runs on Cursor’s own Composer 2.5 and Grok 4.5 models, and is sustainable because owning the model changes the unit economics. That is the whole story.
Cursor Start is the first pricing move Cursor has localized to any country. India is its third-largest market with more than 3 million developers, user base tripled in a year, and the highest power-user concentration anywhere. The plan costs ₹649 per month (~$7), about a third of Pro.
Key takeaways:
- Cursor Start is India-only, ₹649/month, tax-inclusive, billed in INR with UPI or card
- It runs Composer 2.5 and Grok 4.5, not OpenAI or Anthropic frontier models
- Composer 2.5 is third on the Coding Agent Index, near Claude Opus 4.7, at a fraction of the cost
- The price is sustainable because Cursor owns the models, not because it is a loss leader
- You lose Bugbot, Auto Mode, Automations, and the Cursor SDK on Start
How does Cursor Start work?
Here’s how the plan fits. Cursor has three individual tiers. Free lets you try the product with limited agent requests. Pro costs about ₹1,915 per month and unlocks every major model plus the advanced feature set. Start sits between them, aimed at everyday building.
What Start includes:
- Composer 2.5, Cursor’s most powerful model, and Grok 4.5, its most price-efficient coding model
- More agent requests than the free plan, across desktop, web, iOS, and the CLI
- Always-on cloud agents that build, test, and open pull requests while you work
- Cursor for iOS, plus plugins, MCP servers, hooks, and skills
- Local pricing at ₹649 per month, tax-inclusive, payable with UPI
What it drops versus Pro:
- No frontier models from OpenAI and Anthropic
- No Bugbot, Auto Mode, Automations, or Cursor SDK
- India-only access, enforced with VPN blocking and other checks
The comparison table against the other individual plans:
| Plan | India price | Models | UPI | Agent requests |
|---|---|---|---|---|
| Free | ₹0 | Composer | No | Limited |
| Start | ₹649 | Composer 2.5, Grok 4.5 | Yes | Generous |
| Pro | ~₹1,915 | All incl. OpenAI/Anthropic | No | Highest |
Why can Cursor charge a third of Pro?
The short answer: it owns the models. Pro is expensive because renting OpenAI and Anthropic frontier models is expensive. Start runs on Composer 2.5 and Grok 4.5, which Cursor controls and which carry lower operating costs.
The data backs this up. Artificial Analysis puts Composer 2.5 at 62 on its Coding Agent Index, up 14 points from Composer 2. That is third place, behind only Claude Opus 4.7 in Claude Code (66) and GPT-5.5 in Codex (65). On SWE-Bench-Pro-Hard-AA, Composer 2.5 scores 47%, comparable to Claude Opus 4.7 max.
Here’s the cost math that matters:
| Agent | Coding Agent Index | Cost per task |
|---|---|---|
| Claude Opus 4.7 (max) | 66 | $4.10 |
| GPT-5.5 (xhigh) | 65 | $4.82 |
| Composer 2.5 Fast | 62 | $0.44 |
| Composer 2.5 standard | 62 | $0.07 |
Composer 2.5 is about 10 times cheaper than the higher-scoring agents in Fast mode and about 60 times cheaper in standard mode. It sits on the cost-quality frontier: nothing scoring above 60 on the index is cheaper.
The catch: this is the same trick that keeps cheap plans solvent. Renting the most expensive models is what makes low prices lose money. Running on models you control keeps the maths viable at a third of the Pro price. Simon Green, Cursor’s head of Asia-Pacific and Japan, called it “right-sizing the commercial model to drive scale,” not a loss leader.
Why does Composer 2.5 matter for the verdict?
This is the piece that changes the recommendation. When a cheap plan drops to in-house models, you expect a capability step-down. Composer 2.5 is close enough to the frontier that the step-down is small.
It is built on continued training of Moonshot AI’s open-weights Kimi K2.5, with Cursor reporting about 85% of total compute from its own additional training and reinforcement learning. Grok 4.5, SpaceXAI’s model, is explicitly aimed at coding and agentic tasks and posts a SWE-bench Verified score around 86.6%.
That means Start isn’t a “pay less, get a dumber model” plan. It is a “pay a third, keep the near-frontier model, lose the OpenAI and Anthropic options” plan. That is a genuinely different value proposition.
The catch: it isn’t the same as Pro for everyone. If you lean on a specific OpenAI or Anthropic frontier model inside Cursor, Start removes it. And the advanced automation features, Bugbot, Auto Mode, Automations, and the SDK, are gone. For a solo developer doing everyday agentic work, that trade is usually fine. For a team running many parallel cloud agents, Pro still earns its price.
How does Cursor Start fit the India pricing war?
Cursor isn’t first to localize for India, and it isn’t the cheapest. OpenAI launched ChatGPT Go at ₹399 per month in India before taking it global. Anthropic rolled out rupee pricing for Claude this month, but the “localization” is thin: Claude Pro costs about ₹2,000 per month, near the US price.
The landscape right now:
| Product | India price | Notes |
|---|---|---|
| ChatGPT Go (OpenAI) | ₹399 | Sub-$5 tier, consumer-focused |
| Cursor Start | ₹649 | Composer 2.5, Grok 4.5 |
| Cursor Pro | ~₹1,915 | All frontier models |
| Claude Pro (Anthropic) | ~₹2,000 | Near US price, thin localization |
| Claude Max (Anthropic) | ₹11,999 | $125 vs $100 US |
Cursor lands in the middle. Cheaper than its own Pro and than Anthropic, but a real near-frontier coding agent rather than the consumer chatbot tier that ChatGPT Go offers. For Indian developers who build software for a living, that is the relevant comparison.
The catch: this is a fast-moving price war, and the numbers are a snapshot. OpenAI, Anthropic, and Cursor all moved this month. The value of Start could shift as competitors respond.
What is the context around the SpaceX deal?
SpaceX agreed in June to buy Cursor’s parent, Anysphere, for $60 billion in an all-stock deal expected to close in the third quarter of 2026. Cursor operates independently until the deal closes, and Green said the India plans predate the acquisition. SpaceX has been partnered with Cursor on a next-generation coding and knowledge work AI since April.
The timing matters for a different reason. Cursor is discounting in India while The Information reported it is pushing enterprise rates up elsewhere. That isn’t a contradiction. It is a strategy: discount where you buy scale, raise where you have pricing power. The economics of agentic coding are biting across the industry, and GitHub recently froze new Copilot sign-ups when usage costs outran its pricing.
Should you switch to Cursor Start?
For everyday agentic building with a near-frontier model, the math is compelling. ₹649 per month with UPI, tax-inclusive, for a model that scores near Claude Opus 4.7 on coding benchmarks and runs about 60 times cheaper per task. That is a real deal for a solo developer or student.
For power users, it is more nuanced. If you need Bugbot, Auto Mode, Automations, the SDK, or a specific OpenAI or Anthropic frontier model, Start doesn’t cover it. Those features and models are what Pro’s roughly ₹1,915 per month buys. If your work is mostly straightforward agentic coding and you are paying for models you rarely need, Start is worth the switch.
The pattern is the same tiered approach I described in my comparison of AI coding tools: match the plan to the work, not to the maximum capability available. Start covers the everyday 80%. Pro exists for the heavy 20%.
FAQ
What is Cursor Start? Cursor Start is Cursor’s first country-specific plan, built for developers in India. It costs ₹649 per month, tax-inclusive, billed in INR and payable with UPI. It includes Composer 2.5 and Grok 4.5 with more agent requests than the free plan.
How is Cursor Start different from Cursor Pro? Start runs only Cursor’s own models, Composer 2.5 and Grok 4.5, and costs ₹649 per month. Pro costs about ₹1,915 per month and adds frontier models from OpenAI and Anthropic, plus Bugbot, Auto Mode, Automations, and the Cursor SDK.
Is Cursor Start available outside India? No. It is India-only and Cursor uses multiple checks, including VPN blocking, to keep the plan limited to individual users in the country.
Why is Cursor Start so much cheaper than Pro? The plan runs on models Cursor owns rather than renting OpenAI and Anthropic frontier models. Composer 2.5 scores near Claude Opus 4.7 and GPT-5.5 on coding benchmarks at roughly 10 to 60 times lower cost per task.
Is Cursor Start worth it for Indian developers? For everyday agentic building with a near-frontier model, yes. It loses the OpenAI and Anthropic frontier options and advanced automation features. Heavy cloud-agent power users may still need Pro.
Related Posts
- Cursor vs Claude Code vs Copilot: AI coding tools compared: six months of daily use on real coding tasks
- Best AI coding tools for developers in India 2026: pricing in INR and what actually works for Indian developers
- AI tools that accept UPI and Indian payments 2026: which developer tools take UPI, RuPay, and Indian cards
- AI agent cost optimization tips: cut the cost of running AI agents without cutting quality
This article was published on Agentic Up (https://agenticup.dev): practical guides for developers and founders building with AI agents. Reach me at hello@agenticup.dev
