AI coding agents in India: what they really cost
The sticker price is in dollars. The bill arrives in rupees, after conversion, card fees and whatever the exchange rate did that month. Here is the honest comparison.
Monthly cost, in rupees
Approximate public list prices in November 2026, converted at ₹95 to the dollar. Taxes and card fees are extra, and every one of these prices can change without notice.
| Option | Approx. monthly | Payment | Model choice |
|---|---|---|---|
| DoneCorner | ₹200 | UPI, rupees | None — routed for you |
| GitHub Copilot | ≈ ₹950 | USD card | Several models |
| Cursor | ≈ ₹1,900 | USD card | Several models |
| Claude Code | ≈ ₹1,900 | USD card | Anthropic models |
| opencode + your own API | Usage-based | USD card | Any provider |
Competitor prices are public list prices at the time of writing and will change. DoneCorner is ₹200 after the free first month.
The dollar-card problem
Most of these tools assume a US payment method. That means an international card, a bank that allows recurring foreign charges, and a conversion every month. When a card expires or a bank blocks the charge, the agent stops — not because the model failed, but because billing did.
This is the real reason cheap AI tooling has spread slowly in India. It is not only the price; it is the payment rail. A ₹200 UPI charge removes a failure mode that has nothing to do with code.
What ₹200 actually buys
DoneCorner gives you one model name, dc-llm, behind an OpenAI-compatible endpoint. You connect opencode or Cursor and work normally. Behind that name, requests are routed to open-weight models, and the model can change without you changing anything.
- Generous usage for steady daily work — not unlimited.
- No model picker; no chasing releases or price changes.
- Inference served from India and Singapore, so loop latency stays low.
- One active key per account; rotating it revokes the old one.
What it does not buy
- Unlimited tokens. There are daily and monthly caps. Heavy users will hit them.
- Frontier model choice. You get routed open-weight models, not a menu.
- A new editor. It is the model layer; the harness stays yours.
If you need to choose a specific model for a specific task, a direct provider is the better fit. If you want a predictable rupee bill and no model decisions, this is the trade.
Is “free” actually free?
Free harnesses still call a paid model. opencode and Aider cost nothing to install, but every task spends tokens. A handful of long agent sessions can cost more than a monthly subscription — which is why “free tool, paid tokens” is often the most expensive way to start.
The honest summary: the harness can be free, the model cannot. Either you pay per token, or you pay a flat price that includes a cap.
How to start in three steps
- Claim a spot — email and password, no card.
- When approved, create a key in your account.
- Connect opencode with
npx dcorner setup, or Cursor with the base URL and model name.
Questions people ask
Is there a free AI coding agent in India?
The harness can be free: opencode and Aider are open source. The model is not free. You either pay a subscription, pay per token with a card, or use a service like DoneCorner that bundles the model into a rupee price.
Can I pay for an AI coding agent with UPI?
DoneCorner bills in rupees and accepts UPI. Most international tools bill in dollars and expect a credit or debit card enabled for international transactions.
Why is DoneCorner ₹200 when Cursor is around ₹1,900?
Three reasons: it uses open-weight models that cost less to run, it routes each request to a cheaper model when that is enough, and usage is capped rather than unlimited. It is not a launch discount.
Is the usage unlimited?
No. It is generous for steady daily work, but there are daily and monthly caps. We would rather raise a limit than bill you by the token.
Does it work with Cursor and opencode?
Yes. Cursor uses the OpenAI-compatible base URL https://donecorner.ai/v1 with the model name dc-llm. opencode connects with npx dcorner setup.
New to the terminology? Read what a coding harness is first.