Cursor Start vs Pro in India: test the ₹649 plan before treating it as cheap Pro
Quick answer
Cursor Start is a new India-only individual plan priced at ₹649 per month, tax included. It is billed in INR and accepts UPI, credit cards, and debit cards. Start includes access to Cursor's Grok 4.5 and Composer models, more agent requests than the Free plan, always-on cloud agents, iOS access, and extensions such as plugins, MCP servers, hooks, and skills.
It is not a locally discounted version of Cursor Pro. Cursor describes Pro as the plan for developers who need every major model plus Bugbot, Auto mode, Automations, the Cursor SDK, and on-demand usage after included limits. Cursor has not published a precise Start request or token allowance in the launch announcement, so no honest comparison can promise how many projects ₹649 buys.
Choose Start when local billing and daily, bounded work with Grok 4.5 or Composer are enough. Choose Pro when model choice, automation, SDK access, Bugbot, or continued on-demand usage is part of your workflow. Before deciding, run the seven-day accepted-task test below.
Who this is for
This guide is for developers in India searching for Cursor Start pricing, Cursor Start versus Pro, UPI support, or whether the ₹649 plan is useful for real coding work.
It is also for founders choosing an AI coding budget. The useful question is not “Is ₹649 cheap?” It is: How many verified, kept changes does this plan produce for your tasks?
If you already use Auto, read the Cursor Router rollout guide. If you are evaluating Grok itself, reuse the fixtures in the Grok 4.5 coding-agent evaluation.
What changed
Cursor launched Start on July 28, 2026 for developers in India. The plan sits between Free and Pro. Cursor says its Indian user base has grown to more than three million developers and that local users asked for regional pricing and UPI payment.
These are the confirmed plan boundaries from Cursor's announcement and current pricing page:
| Decision point | Free | Start | Pro |
|---|---|---|---|
| Price | No payment required | ₹649/month, tax included | $20/month before applicable taxes |
| Availability | General entry plan | Developers in India | Standard individual plan |
| Payment | None | INR via UPI or card | Major credit and debit cards |
| Models described by Cursor | Composer with limited local agent requests | Grok 4.5 and Composer | Grok, Composer, and frontier models from other labs |
| Agent capacity | Limited | More requests than Free; exact allowance not published in the launch | Extended limits plus optional on-demand usage |
| Cloud agents | Not promised in the Free description | Included | Included |
| Plugins, MCP, hooks, skills | Not promised in the Free description | Included | Included |
| Bugbot, Auto, Automations, SDK | Not included in Cursor's Start list | Cursor positions these as Pro capabilities | Included or available with Pro |
The wording matters. “More” and “generous” are relative descriptions, not a public request count. Do not translate them into tokens, agent hours, or completed repositories without account-level evidence.
Choose by workload, not sticker price
Stay on Free
Use Free to learn the interface and run three small fixtures: one explained code change, one tested bug fix, and one short refactor. Upgrade only after the workflow itself saves time.
Choose Start
Start fits developers in India who can keep most work inside Grok 4.5 and Composer, want UPI or INR billing, and run bounded tasks with deterministic checks. Examples include a small feature with tests, a focused refactor, documentation tied to code, or launching a cloud agent and reviewing its pull request.
Start becomes a weak fit when your standard workflow requires a particular third-party model, Auto routing, scheduled Automations, the SDK, Bugbot, or predictable continuation after the included allowance is consumed.
Choose Pro
Choose Pro when model diversity is a requirement rather than a preference, or when one of the Pro-only workflow surfaces is part of the job. Pro is also the clearer path for heavy use because Cursor explicitly documents on-demand usage after included model usage. That does not make Pro automatically cheaper; it makes its continuation path clearer.
Run a seven-day accepted-task test
Do not spend the week “vibe coding” one large app. Build a small, repeatable evaluation pack.
1. Select 12 real tasks
Use four task lanes with three tasks each:
| Lane | Example | Required verification |
|---|---|---|
| Small edit | validation rule, copy change, one test | focused test and diff review |
| Product change | bounded feature across a few files | typecheck, tests, build |
| Diagnosis | reproduce and fix a known bug | failing-before/passing-after evidence |
| Cloud-agent work | open a draft PR from a clear issue | current-commit checks and PR review |
Keep the repository, instructions, permissions, and verification commands fixed. Do not include a security-critical autonomous repair unless a human can independently review it.
2. Record allowance movement
For each task, record the visible plan allowance before and after, correction turns, elapsed time, verification result, and whether you kept the change 24 hours later. Because Cursor does not publish a Start token quota, use the account's visible percentage or usage unit without pretending it is a token count.
3. Calculate accepted-task cost
Use two simple metrics:
allowance_per_accepted_task = total_allowance_used / accepted_tasks
cash_per_accepted_task = monthly_plan_price / expected_monthly_accepted_tasks
A low request cost is meaningless if the patch is rejected or requires a second tool to redo it. Count a task only when its tests pass, the diff survives review, and you would keep it.
4. Apply stop gates
Stop the Start trial and reconsider the plan when any of these happens:
- a medium task consumes more allowance than your monthly task budget permits;
- the agent claims a fix that deterministic verification disproves;
- your normal work repeatedly requires a model or feature reserved for Pro;
- corrections erase the time saved;
- a cloud-agent result cannot be tied to the reviewed commit.
Use the iPad cloud-agent PR checklist if you review away from desktop. Keep the untrusted-repository sandbox gate regardless of plan.
Copyable evaluation record
cursor_plan_eval:
plan: "start"
region: "IN"
price_inr: 649
period: "2026-08-01..2026-08-07"
task_id: "product-03"
task_lane: "product-change"
model: "grok-4.5 | composer"
allowance_before: ""
allowance_after: ""
correction_turns: 1
verification:
tests: "passed"
build: "passed"
human_review: "accepted"
kept_after_24h: true
elapsed_minutes: 34
requires_pro_feature: false
Review all 12 records together. Stay on Start only if the accepted-task rate, remaining allowance, and missing-feature count fit a normal month.
Common mistakes
Treating Start as ₹649 Pro. The cheaper plan intentionally has a narrower model and feature boundary.
Inventing a quota. Cursor's launch post does not give an exact Start request or token allowance. Account observations can inform your own test, not a universal promise.
Comparing only monthly prices. Compare cost per accepted change, correction time, and failed-task severity.
Using a VPN as an eligibility strategy. Start is announced for developers in India. Plan access and billing eligibility are product boundaries, not an arbitrage checklist.
Trusting autonomous output without evidence. Lower price does not lower the cost of an incorrect security, payment, or data change.
FAQ
Is Cursor Start available outside India?
Cursor announced Start for developers in India and prices it in INR with UPI or card payment. The official announcement does not promise availability in other countries.
Does Cursor Start include Claude or OpenAI models?
Cursor lists Grok 4.5 and Composer for Start. It describes Pro as the plan with access to every major model, including advanced models from other labs. Buy Start only if the listed Cursor models cover your workload.
How many agent requests does Cursor Start include?
Cursor says Start has more agent requests than Free and calls its access generous, but the launch page does not publish a precise request or token number. Use the seven-day test and your account's visible usage data.
Is Cursor Start worth ₹649?
It can be when INR/UPI billing matters and the included models complete enough verified tasks. It is not a good substitute for Pro when you need model choice, Auto, Automations, the SDK, Bugbot, or on-demand continuation.
Top comments (0)