Home / Nexa Quant / Algo trading in India: what the 2025 retail API rules actually changed

Algo trading in India: what the 2025 retail API rules actually changed

Website Development✳App Development✳UI/UX Development✳Nexa Flux✳Nexa Quant✳Content Creation✳
Content Creation✳Nexa Quant✳Nexa Flux✳UI/UX Development✳App Development✳Website Development✳
Algo trading in India: what the 2025 retail API rules actually changed

Algo trading in India: what the 2025 retail API rules actually changed

28 September 2026

If you automate orders through a broker API in India, the rules you operate under changed meaningfully. Most of the commentary has been either alarmist or vague, so this is the operational version: what you have to do differently, and what genuinely has not changed.

Registration and tagging

The headline change is traceability. Orders placed through an API above a defined frequency threshold are expected to be tagged and identifiable, with the strategy registered through the broker and the exchange. In practice this means your execution layer needs to attach a consistent identifier to every order it sends, and you need the paperwork behind that identifier to exist.

This is not difficult to build. It is difficult to retrofit. If your automation currently sends orders indistinguishable from manual clicks, adding identifiers after the fact usually means touching every code path that places an order.

Where the frequency threshold bites

  • Low-frequency, discretionary-assist automation is largely unaffected in day-to-day operation.
  • Strategies that fire frequently inside a session fall under the stricter end of the framework.
  • Order modification and cancellation count toward activity, which catches a lot of options strategies that re-quote often.

The practical answer is to instrument your system so you can measure your own order rate per second and per minute before anyone asks. Most desks we onboard have never measured it.

What did not change

You still trade your own capital, in your own broker account, with keys you create and can revoke. Software vendors do not take custody. Nobody can promise returns, and any vendor showing you a backtest without brokerage, STT and slippage applied is showing you fiction.

The useful mental model: the regulator is not trying to stop you automating. It is trying to make sure that when something goes wrong at 2:30pm, the order trail explains itself.

A sensible checklist

  • Tag every order with a stable strategy identifier.
  • Keep immutable logs of signals, orders, fills and rejections.
  • Implement a hard kill switch independent of strategy logic.
  • Reconcile your own trade log against the broker contract note daily.
  • Document your strategy rules in writing before you automate them.

None of this is exotic. It is the same discipline an institutional desk has run for decades, scaled down. The desks that treat it as a compliance tax tend to be the ones who find out the hard way that their logs were incomplete.