Algo trading in India: what is allowed, what it costs, what breaks.

01Is it legal

Yes. Algorithmic trading has been permitted in India for many years, initially for institutional participants and subsequently for retail participants through broker-provided automation. The framework is set by SEBI and administered through the exchanges and your broker.

The direction of regulation has been consistently toward more accountability rather than less. The general shape is that algorithms made available to retail investors are expected to be registered with the exchange through the broker and to carry a unique identifier, so that orders generated by an algorithm can be attributed to it. Brokers carry obligations around what they permit, how they monitor order flow, and what they must disclose.

02When registration matters

The most common confusion we encounter is the belief that writing trading software requires SEBI registration. It does not. Registration requirements attach to activities:

The line is whether you are acting on someone else's financial decisions or merely building a tool they operate. This is precisely why QuantNifty builds and hands over systems rather than running client money or selling subscriptions: it keeps us on the engineering side of that line. We are not SEBI registered, and we do not give advice.

03What a broker API gives you

Every major Indian broker with an automation offering exposes broadly the same capability set:

• Order operations: place, modify and cancel orders across product types, including market, limit and stop-loss variants.
• Position and margin state: current holdings, open positions, margin used and margin available.
• Streaming market data: live quotes and depth over a websocket, usually with a cap on how many instruments you may subscribe to at once.
• Historical data: candles at various intervals, with limits on how far back and at what granularity.
• Postbacks or order updates: asynchronous notification that an order's state has changed.

That is genuinely enough to run a serious systematic strategy. The gap between "enough" and "reliable" is where the engineering lives.

04What it does not give you

Assumptions that a retail API quietly violates, in rough order of how expensive they are to learn the hard way:

• That a placed order is a filled order. It is not. If your code assumes the position changed the moment the order was accepted, your position size will drift away from what your strategy believes it is holding. Every order needs a confirmation path.
• That your connection will stay up. It will not. If a stop-loss exists only as logic in your process, a disconnect leaves a short option position completely unprotected. Stops for short-option strategies belong on the broker's book.
• That the quote you saw is the price you get. On an illiquid strike or during a momentary liquidity hole, a market order can fill far from the last traded price. This is what spike detection and order repricing exist to prevent.
• That rate limits will not bite. They will, usually at the worst moment, when the strategy is trying to exit several legs at once.
• That you can intervene manually in time. You cannot. A kill switch that flattens positions and halts trading on a defined condition needs to exist in the system and be reachable without logging into a terminal.

05The real cost structure

Running costs split into two very unequal halves.

Fixed and recurring

• Broker API subscription, typically a modest monthly fee per app.
• A server, ideally hosted close to the exchange to keep round-trip latency low and stable.
• Historical data for research. Tick-level option data is the expensive item, and it is not optional if you intend to test intraday option strategies honestly.

Transactional, and much larger

Brokerage, exchange transaction charges, Securities Transaction Tax, stamp duty, SEBI turnover fees and GST, plus slippage on every entry and every exit. For any strategy that trades frequently, this category dwarfs everything else.

06Infrastructure you need

A production setup for an intraday index options strategy, at minimum:

• A server that starts the strategy automatically before market open and does not depend on your laptop being awake.
• State that survives a restart, so a process crash mid-session does not lose track of open positions.
• Logging detailed enough to reconstruct exactly why a trade happened, months later.
• Alerting that reaches you on a phone when something needs a human.
• A reconciliation step at end of day comparing what the strategy thinks it did against what the broker says happened. Discrepancies here are how you find bugs before they compound.

07Where retail systems break

Across the systems we have been asked to rescue, the failures cluster tightly:

• Backtest ignores charges, so a strategy with a genuine but thin edge is deployed and bleeds out through costs.
• Stop-loss lives in the script, and one disconnect during a volatile session produces a loss larger than a year of gains.
• No expiry calendar, so holiday-shifted expiry weeks trade the wrong contract.
• Overfitting through repeated tweaking: the strategy was adjusted after seeing out-of-sample results, which quietly converts the out-of-sample period into an in-sample one. See honest backtesting.
• Position sizing set by what the margin allows rather than by what the drawdown tolerance permits.

08Questions people ask