Is Algo Trading Legal in India? SEBI Rules, AI Bots and Red Flags
Yes, algo trading is legal in India. SEBI's retail algo rules apply to all brokers from 1 Apr 2026. What is allowed, AI bots, red flags and a 10-point check.
Yes. Algo trading is legal in India, for retail traders as well as big firms. SEBI (the market regulator) set a retail algo framework in a circular dated 4 Feb 2025. A later circular made it apply to every stock broker from 1 Apr 2026. An algo here means orders made by computer rules instead of by your own clicks.
Legal does not mean "anything goes". Algo orders must go through your broker's approved API (the door that lets software send orders), with an exchange tag on each order. Algo sellers must be listed with the exchanges. Anyone who sells a secret strategy must be a SEBI registered research analyst (a SEBI licence for people who give stock research). A scheme that offers fixed or sure returns is not allowed, whatever software it uses.
The same rules cover "AI trading bots". The SEBI and NSE algo documents we read have no separate category for AI. What matters is who places the order, through which route, and who is responsible for it. This guide explains the rules in plain words and maps them to four common situations. It works through the 10 orders per second rule. It ends with a 10-question check for any AI or algo tool.
Key Takeaways
- 1.Algo trading is legal for retail traders in India. SEBI's framework of 4 Feb 2025 applies to all stock brokers from 1 Apr 2026.
- 2.NSE counts every order you send through a broker API as an algo order. Each one carries an exchange tag.
- 3.Your own code needs exchange registration only if it sends more than 10 orders per second per exchange. A registered algo may be used only for you and your family.
- 4.SEBI does not register algo software. Exchanges empanel (approve and list) algo providers, and black box providers must also be SEBI research analysts.
- 5.An "AI bot" follows the same rules as any algo. A promise of fixed or sure returns is a red flag, and NSE says such schemes are prohibited by law.
Is algo trading legal in India? The short answer
Yes, with rules. SEBI first set broad algo guidelines for the market in a circular of 30 Mar 2012. For years, algos were used mostly by institutions through direct market access. This facility lets big investors send orders to the exchange through their broker's systems. The 4 Feb 2025 circular, titled "Safer participation of retail investors in Algorithmic trading", opened a clear path for retail traders through their brokers.
If you want the basics first, our page on what algo trading is explains how a rule turns into an order. This page sticks to the legal question.
What counts as an algo order
SEBI's circular describes an algo as "orders generated using automated execution logic". NSE goes further in its retail algo FAQ. It says all orders received from clients through an API are treated as algo orders and need the right tag. This is true even when your orders stay within the threshold of 10 orders per second.
So a small script that places one trade a day is still an algo in the eyes of the exchange. Orders you place by hand on your broker's app or website are not.
What stays illegal, with or without an algo
- Schemes with sure returns. NSE's caution of 13 Oct 2026 tells investors not to subscribe to schemes offering returns which are indicative, assured or guaranteed. It says such schemes are prohibited by law.
- Account handling. The same NSE caution was about a Telegram channel offering "account handling services". NSE tells investors not to share trading credentials such as user ID and password with anyone.
- Selling a hidden strategy without registration. A black box algo provider must register with SEBI as a research analyst.
- Lending your registered algo to friends. SEBI allows a retail trader's own registered algo to be used for family, but not for other investors.
- Open APIs and shared logins. SEBI tells brokers not to permit open APIs. Access is only through a unique API key and a static IP the broker has whitelisted.
SEBI algo trading rules in plain words
The framework gives jobs to four groups: SEBI, the stock exchanges, your broker and the algo provider. You, the investor, sit at the end of the chain. Here are the dates first, because many older articles still show the first start date.
The dates that matter
- 4 Feb 2025: SEBI issues circular SEBI/HO/MIRSD/MIRSD-PoD/P/CIR/2025/0000013. It was first due to apply from 1 Aug 2025.
- 5 May 2025: NSE publishes its implementation standards in circular NSE/INVG/67858. This holds the static IP and 10 orders per second rules.
- 29 Jul 2025: SEBI moves the start to 1 Oct 2025, as its later circular records.
- 30 Sep 2025: SEBI circular SEBI/HO/MIRSD/MIRSD-PoD/P/CIR/2025/132. Brokers who are ready go live from 1 Oct 2025. Others get a glide path (a set of step dates).
- 31 Oct 2025, 30 Nov 2025 and 3 Jan 2026: the glide path milestones. Brokers had to apply for and then finish registering their retail algo products, and join a mock session.
- 5 Jan 2026: brokers who miss the milestones are barred from onboarding new retail clients for API algo trading.
- 1 Apr 2026: the framework, with the exchanges' standards, applies to all stock brokers.
Why do some people think algo trading was banned? The start date moved twice, from 1 Aug 2025 to 1 Oct 2025, and then to 1 Apr 2026 for all brokers. SEBI also barred brokers who missed the milestones from onboarding new API algo clients from 5 Jan 2026. Some tools that could not meet the rules stopped working. None of this made algo trading illegal. It made it traceable.
Who does what under the framework
| Who | Main duty | Where it is written |
|---|---|---|
| SEBI | Sets the framework. Does not regulate algo providers directly | SEBI circular, 4 Feb 2025, part III |
| Stock exchange (NSE, BSE) | Empanels algo providers, registers algos, gives algo IDs, watches all algo orders, can use a kill switch (an instant stop) on a rogue algo | SEBI circular part IV; NSE/INVG/67858 |
| Your broker | Acts as the principal (the party responsible). Gets exchange approval for each algo, whitelists static IPs, rejects orders above the limit, handles complaints | SEBI circular parts I and II; NSE/INVG/67858 |
| Algo provider (vendor or platform) | Acts as the broker's agent. Must be empanelled with each exchange and register its algos | SEBI circular parts I and III; NSE/INVG/67858 |
| Black box algo provider | Also registers with SEBI as a research analyst and keeps a research report for each algo | SEBI circular part V; NSE retail algo FAQ |
| You, with your own code | Give the broker a static IP, stay at or under 10 orders per second per exchange or register the algo, log in again each day | NSE/INVG/67858; NSE retail algo FAQ |
White box and black box algos
SEBI sorts algos into two groups. A white box algo, also called an execution algo, is one where the "logic is disclosed and replicable". You can see the rules and copy them. A black box algo is one where "the logic is not known to the user and is not replicable". You only see the trades. That second group carries extra duties: the provider must be a SEBI research analyst and keep a research report for each algo. If the logic changes, the algo must be registered again as a new one.
Static IP, two-step login and daily logout
If you run your own code, NSE's standards say you must give your broker a static IP (a fixed internet address). You can change it, but not more than once a calendar week, except in rare cases agreed with your broker. One static IP maps to one client, though a family may share one. All API sessions must be logged out every day before the next trading day starts.
SEBI also tells brokers to use OAuth login only for API access. OAuth lets an app in without you sharing your login details. Brokers must also use two factor authentication (a second check, such as an OTP). Do you use a vendor's platform hosted on your broker's servers? Then NSE's FAQ says the broker server's static IP is used, so you do not need your own. The same FAQ quotes an NSE rule: algo orders with order type as market order are not permitted. So a script should send limit orders (orders with a set price). Our guide to static IP for algo trading covers how to set one up.
What you can do as a retail trader: four situations
The rules make more sense once you match them to what you actually do. Most retail traders fall into one of these four cases.
| What you do | Is it allowed? | What you need | Who answers for it |
|---|---|---|---|
| Place every order yourself in your broker's app | Yes. This is not algo trading | Nothing extra | You |
| Run your own script (Python or similar) on your broker's API | Yes, through the retail framework | API key, static IP, daily login. Register the algo only above 10 orders per second | Your broker for the order route and risk checks; you for the logic and the result |
| Use a no-code or vendor platform | Yes, if the provider is empanelled | An empanelled provider whose algos have exchange IDs and run on your broker's servers | Your broker as principal, the provider as its agent |
| Buy signals or a ready "AI bot" strategy | Only through registered parties | For a hidden strategy: a SEBI research analyst who is also an empanelled algo provider | The registered provider and the broker |
Your own code and your family. SEBI says an algo you build yourself may be used for your family but not for other investors. Family here means self, spouse, dependent children and dependent parents. NSE's FAQ adds that a tech-savvy client who runs their own logic carries its profit or loss. The broker is responsible for risk checks.
No-code platforms. NSE's FAQ says all algos from algo providers must be hosted on the broker's server. That is why a vendor platform now runs through your broker's systems. Our Tradetron, Streak and AlgoTest comparison shows how three such platforms differ, and how to start algo trading in India walks through the first steps.
Does Zerodha allow algo trading? Zerodha's developer docs describe Kite Connect as "a set of REST-like HTTP APIs" that let you place orders and manage a portfolio. Other brokers have their own APIs. Whichever broker you use, read its own API page and ask whether your setup is live under the retail algo framework. Our guide on choosing a broker API lists what to compare. We do not rank brokers.
The 10 orders per second rule, with a worked example
This is the line that decides whether you must register your own algo. NSE's standards set the Threshold Order Per Second (TOPS) at not more than 10 orders per second per exchange. Below it, you do not need to register your algo. Your orders still carry a generic exchange tag. Above it, you must register the algo with each exchange through your broker, and the orders then carry the exchange's algo ID.
Three more details matter. The limit is "initially set" at 10, and NSE says exchanges may change it after notice. It is measured by the calendar clock second of the broker's server. And if your orders cross the limit without registration, the broker shall reject or not process the orders above it.
- Example script A sends one order every 2 seconds. That is 1 ÷ 2 = 0.5 orders per second, far below 10. No registration is needed, but each order is still tagged as an algo order.
- Example script B fires a basket of 12 orders inside one clock second on the same exchange. That is above 10. Without a registered algo, the broker would reject the orders above the limit in that second.
- At exactly 10 orders in every second, a script would send 10 × 60 = 600 orders in a minute. Very few retail strategies need anything close to that.
- Example script C places 40 trades a day, each with an entry and an exit. That is 40 × 2 = 80 orders across a whole session, well below the limit in any one second.
All numbers in this list are made-up examples. The point: most retail scripts never come near the threshold. The real risk is a loop bug that fires many orders at once. Your broker may also set its own lower limit per client, which NSE's standards allow.
Is AI trading legal in India?
Yes, if the orders follow the algo rules above. The SEBI circulars and NSE standards we read do not create a special category for AI. They look at how an order is made and sent. An "AI" label does not make a tool legal or illegal.
AI trading bots
An AI trading bot in India is just an algo that uses some learning model to decide its trades. If it sends orders through your broker's API, it is an algo order. If the seller hides its logic from you, it is a black box. For a black box algo, SEBI says the provider shall "Register as a Research Analyst and maintain a detailed research report for each such algo". If the logic changes, it must be registered again as a fresh algo. An "AI bot" seller who has neither exchange empanelment nor research analyst registration is outside these rules.
Using ChatGPT, Claude or Gemini to write your script
Writing code with a chat assistant is ordinary software use. Once that code places orders through your broker's API, the usual rules apply. That means a static IP, daily login, tags and the 10 orders per second line. The responsibility for the logic stays with you. Chat models can also invent an old lot size or a wrong function name, so test the code before it touches real money. Our page comparing ChatGPT, Claude and Gemini for trading shows where each one slips.
AI apps that only give ideas
Many AI tools never place an order. They read filings, describe charts, answer questions or review your past trades. These are not algos. But an app that tells you what to buy or sell is giving advice. Advice for a fee generally needs SEBI registration, for example as an investment adviser or research analyst. Our AI trading guides sort these tools by the job they do. If you plan to let an AI agent act on a live broker account, read AI agents placing trades first.
Who is registered with whom
Search results often talk about "SEBI registered algo software". That phrase is misleading. SEBI's circular says algo providers "shall not be regulated by SEBI". Instead, any provider that places algo orders through a broker's API must be empanelled with the exchanges. Here is the chain in plain words.
- SEBI registers brokers and research analysts. It does not register algo software.
- Exchanges empanel algo providers and register each algo with a unique ID. For example, NSE circular NSE/INVG/76481 of 22 Sep 2026 gave final empanelment to 7 providers that were earlier empanelled on a provisional basis. Its table shows each provider's algo type, white box or black box.
- Brokers must deal only with empanelled providers, check them before signing up, and tell the exchanges about each arrangement.
- Black box providers need both: exchange empanelment and SEBI research analyst registration.
So do not ask "is this app SEBI registered?" Ask two other questions. Is this provider on the NSE or BSE empanelled list? Does my broker work with it? Our page of empanelled algo providers collects the exchange lists with dates.
Red flags in AI and algo trading products
Most trouble does not come from the rules. It comes from people selling around them. These warning signs come straight from official notices.
| Red flag | What the official text says | Source |
|---|---|---|
| Fixed, assured or sure monthly returns | Do not subscribe to schemes offering returns which are indicative, assured or guaranteed; such schemes are prohibited by law | NSE caution, 13 Oct 2026 |
| Asks for your password, PIN or OTP | Do not share trading credentials such as user ID and password; brokers must use OAuth and two factor login for APIs | NSE caution; SEBI circular, 4 Feb 2025 |
| Offers to handle your account | NSE warned about a Telegram channel offering account handling services | NSE caution, 13 Oct 2026 |
| Live trading tips on social media | Do not trust the claims of such persons; deal only with SEBI registered intermediaries | SEBI press release, 17 Aug 2026 |
| Secret strategy, no research analyst number | Black box algo providers must register as research analysts | SEBI circular, 4 Feb 2025 |
| Not on any exchange list | Algo providers must be empanelled with each exchange where their algos trade | NSE/INVG/67858 |
| Money sent to a personal account | Brokers' designated client bank accounts are shown on NSE's broker search page | NSE caution, 13 Oct 2026 |
Two more signs need no notice. A win rate with no average loss tells you little, because a few big losses can wipe out many small wins. And a record that shows only winning screenshots is not a record. A real one lists every trade with dates, losers included.
10 questions to ask before you use any AI or algo tool
Run these on any tool, free or paid, before you give it money or access. If you cannot get a clear answer, treat that as the answer.
| # | Question | A good answer |
|---|---|---|
| 1 | Does it place orders, or only give information? | It says plainly which one |
| 2 | Which broker does it work through? | Named brokers, and it runs on the broker's servers |
| 3 | Is the provider empanelled with NSE or BSE? | You can find it on the exchange list |
| 4 | Is the logic shown to you (white box) or hidden (black box)? | If hidden, a SEBI research analyst number you can check |
| 5 | Does each algo have an exchange algo ID? | Yes, and your broker can confirm it |
| 6 | Does it promise any return? | No. It shows risk warnings instead |
| 7 | Does it ask for your password, PIN or OTP? | Never. Login is through your broker's own screen |
| 8 | What are all the charges? | Subscription and brokerage shared with the broker are fully disclosed, as SEBI requires |
| 9 | How are past results shown? | Every trade, average win, average loss, charges and the test period |
| 10 | What happens if it goes wrong? | A clear stop, a daily limit and a named person for complaints, usually at your broker |
Question 8 comes from the circular itself. SEBI lets algo providers and brokers share subscription charges and brokerage. But it says "prominent and complete disclosures of all the charges shall be made to the client". NSE adds that brokers may charge API fees on top of brokerage, including exchange charges for algo registration.
Do algos make money? What SEBI's data says
Legal is one question. Profitable is another. SEBI's press release of 20 Aug 2026 and its study of the same month give the clearest Indian numbers.
- 87.7% of individual traders in equity derivatives (futures and options) made a net loss in FY26.
- Their total net loss was about ₹91,685 crore in FY26, compared with about ₹1.12 lakh crore in FY25.
- Individual traders paid transaction costs of around ₹25,000 crore in FY26.
- 99% of the gross profits of FPIs (foreign portfolio investors) and proprietary traders (firms trading their own money) were made by "algo entities". SEBI counts an entity as an algo entity if it made at least one algo trade in the year.
This does not mean "use an algo and you will profit". The study shows who uses algos. In FY26, 372 of 432 FPIs and 323 of 693 proprietary traders traded through algos. Among individuals, only 15% of 84.2 lakh traders did, and SEBI says even this "overstates genuine algorithmic activity". The reason: it includes traders whose intraday positions were squared off (closed) automatically by their brokers. SEBI's earlier study gave similar figures for FY24: 97% of FPI profits and 96% of proprietary profits came from algo entities.
Big firms win with scale, speed and very low costs per trade. A retail script has none of these. What it does have is the same charges on every order as a manual trader. Our page on algo vs manual trading compares the two in more depth.
Worked example: charges on a small edge
Here is a made-up example of a bot that buys and sells NSE options. It makes a small gross gain on each trade. The prices and the ₹20 brokerage per order are examples. The other rates are the statutory rates in force on 10 Nov 2026.
- Example trade: buy options for a premium of ₹10,000 and sell them for ₹10,100. Gross gain: ₹100.
- Brokerage at an example ₹20 per order, two orders: ₹40.
- STT (securities transaction tax) at 0.15% of the sell premium: ₹10,100 × 0.15% = ₹15.15.
- NSE transaction charge at 0.03553% on both sides: ₹20,100 × 0.03553% = ₹7.14.
- SEBI fee at ₹10 per crore of turnover: ₹0.02.
- GST at 18% on brokerage, exchange charge and SEBI fee: ₹8.49. Some brokers leave the SEBI fee out of the GST base; here that changes less than one paisa.
- Stamp duty at 0.003% of the buy premium: ₹0.30.
- Total charges: ₹71.10. Net gain per trade: ₹100 minus ₹71.10 = ₹28.90.
Now say the bot does this 40 times a day. Gross gain: ₹4,000. Charges: ₹2,844. Net: ₹1,156. Charges ate 71.1% of the gross gain, and that is before slippage (getting a worse price than expected) or any losing trades. A bot with a small edge per trade can look busy and still keep very little. Our page on algo trading charges adds platform and API fees on top of these.
Keep a journal of your algo and manual trades
Whether you click every order or run a script, one question decides if it is worth it: what is left after charges? A broker's P&L screen will not tell you whether your algo days beat your manual days, or which rule change helped. A trading journal can.
- Log each trade with a tag such as "manual" or "algo", plus the strategy or script version.
- Note every change you make to a bot's rules, with the date and the reason.
- Write down when you switched a bot off or raised its size, and why. These are the same emotional choices a manual trader makes, just one step earlier.
- After a set number of trades, compare the groups on net P&L, win rate, average loss and drawdown (the drop from your highest balance).
OneTradeJournal is a manual trading journal built for Indian F&O traders. You enter each trade by hand, so it suits this kind of tagging. It estimates brokerage and statutory charges per trade, so your P&L shows after charges, and it reports win rate, profit factor and drawdown by strategy. It is not an algo or signal tool. It places no orders, does not link to your broker, and is not SEBI registered. You can start on the free plan, which lets you log one trade a day in each market. AI review features such as Ask Your Journal sit in the paid plan, and the AI only knows what you logged. Our AI trading journal guide shows how that review works.
Sources
- SEBI circular SEBI/HO/MIRSD/MIRSD-PoD/P/CIR/2025/0000013, Safer participation of retail investors in Algorithmic trading, 4 Feb 2025, checked 10 Nov 2026
- SEBI circular SEBI/HO/MIRSD/MIRSD-PoD/P/CIR/2025/132, extension of timeline, 30 Sep 2025, checked 10 Nov 2026
- NSE circular NSE/INVG/67858, implementation standards for retail algo trading, 5 May 2025, checked 10 Nov 2026
- NSE, FAQs on safer participation of retail investors in algorithmic trading, checked 10 Nov 2026
- NSE circular NSE/INVG/76481, final empanelment of algo providers, 22 Sep 2026, checked 10 Nov 2026
- NSE press release, Caution for Investors, 13 Oct 2026, checked 10 Nov 2026
- SEBI press release on trading by individuals in equity derivatives, 20 Aug 2026, checked 10 Nov 2026
- SEBI study, Profitability of Individual Traders in the Equity Derivatives Segment (FY25 to FY26), Aug 2026, checked 10 Nov 2026
- SEBI study, Analysis of Profits and Losses in the Equity Derivatives Segment (FY22 to FY24), Sep 2024, checked 10 Nov 2026
- SEBI press release, caution on live trading strategies on social media, 17 Aug 2026, checked 10 Nov 2026
- NSE circular NSE/FATAX/73524, changes in rates of STT, 31 Mar 2026, checked 10 Nov 2026
- Zerodha, Kite Connect developer documentation, checked 10 Nov 2026
Related Topics
Related Articles
Paper Trading App in India: What Each App Offers and What SEBI Says
Paper trading apps in India compared from their own pages: cost, virtual money and price delay. Plus SEBI's 2024 advisory and how to judge your practice trades.
Share Market Kaise Sikhe: A Step by Step Plan for Beginners in India
Share market kaise sikhe: a 12-week beginner plan, free NISM and NSE resources, real costs of a first trade, SEBI loss data and a journal habit. No promises.
TradingView Paper Trading in India: Steps, Limits and Nifty Rules
TradingView paper trading is free, but its help page lists Indian exchange symbols as non-tradable. Setup steps, NSE data delay, limits and a journal method.
Zerodha Console P&L: How to Read Your Report and What It Misses
Read your Zerodha Console P&L step by step: realised and unrealised P&L, charges, heat map, Tax P&L, Verified P&L, why numbers differ and what it can't show.
TradeZella Price in India: Plans in Dollars and Rupees (2026)
TradeZella price: Essential $35, Pro $59, Ultra $99 a month or $315 to $891 a year. Rupee cost, the July 2026 rise, trial facts and checks for Indian traders.
Trade Book Meaning: What Your Tradebook Shows and How to Use It
A trade book lists only your filled trades. See how it differs from the order book and where Zerodha, Groww, Angel One, Dhan and Upstox keep it.
The trading journal for Nifty, Bank Nifty and Indian F&O. Track your trades, spot patterns, build discipline.
- Log one trade a day by hand, on purpose
- AI mentor finds your repeat mistakes
- Behavioural analytics catch tilt early
- Trading calendar with P&L heatmap
- Pre-trade checklist flags risks
Yearly ₹3,999 · No broker credentials