A complete algorithmic trading terminal for Indian markets. Describe a strategy in plain conditions — no code. Test it against years of real NSE and BSE history on a Rust engine that charges you for every rupee of brokerage, STT, GST, stamp duty and slippage. Then deploy it to paper or to your own broker account.
No card required · Every account is reviewed and approved by our desk
A rule that looks brilliant on a chart you scrolled by hand is a different animal once it pays brokerage on every leg and gets filled a tick worse than you assumed. GTP Algo is built to find that out before your capital does.
Not a screener bolted to a chart. A research environment, an execution surface and the data layer under both — built together, so a number never means two things.
Every fill is charged brokerage, STT, exchange transaction charges, GST, stamp duty and your configured slippage. The headline is what is left — not the gross figure that flatters every intraday system ever written.
A native engine with a per-core worker pool sitting directly on columnar bar storage. No interpreter, no queue, no notebook. You iterate at the speed you think.
Conditions are built from dropdowns — indicator, source, comparison, offset — and read back like a sentence. Nothing to learn, nothing to debug, and no class of bug where the script does something other than what you meant.
Real setups happen in stages. Flows let one step wait on another across different timeframes, with minimum and maximum bar gates and ALL / ANY / QUORUM joins — so "the daily holds, then the 5-minute breaks" is one strategy, not two.
Argon2id passwords, single-use rotating refresh tokens with theft detection, ticketed WebSockets that keep your token out of every proxy log, and hardware-backed storage on Android.
Years of NSE and BSE history, minute and daily, kept synced and ready — nothing for you to buy, plug in or keep alive. Every index resolves to its constituents at run time, so a strategy that trades "the Nifty 50" follows a reconstitution on its own. You open a chart or start a backtest and the bars are simply there.
Nothing here is a mock-up or a roadmap slide. Pick a module and read exactly what it does, what each control changes, and why it was built that way.
Up to eight synchronised panes in one grid, a docked watchlist, and a bar replay that walks history forward one candle at a time so you can test your own eye against the same data your strategy sees.
Close, Open,
High, Low, HL2 (median),
HLC3 (typical) or OHLC4 — the same seven the strategy
builder offers, so a study you tuned on a chart behaves identically in a rule.
The builder walks you through every decision a mechanical strategy has to make — and refuses to let you skip one. Nothing saves itself behind your back: you edit, you review, you press Save, and Discard genuinely discards.
Intraday (squared off at the session's square-off time, nothing
carried overnight), Swing (carried overnight, force-closed on the
horizon you set in Schedule), or Long term (nothing forces an
exit — only your own rule, stop or target does).
live
(inside a referenced index or watchlist, re-resolved when the strategy runs, so
it follows a reconstitution), picked (named by hand, frozen), and
excluded (subtracted after every source resolves — it beats all of
them). A left rail gives the category-wise count, a grid gives the symbol-wise
answer.
[-1] reach back in history, and
[0] reads the forming candle — evaluated minute by minute
rather than only at the close.
Fixed capital — a set rupee amount per trade, with quantity derived
from the entry price — or Fixed quantity — the same share count
every trade, whatever the price. Both numbers are kept when you switch, so
changing your mind never destroys the value you had under the other method.
Slippage and commission live here too, because a scalper on
illiquid midcaps and a positional strategy on Nifty 50 do not share a cost model.
/strategies/{id} opens that one in the builder, so a half-written
idea is a link you can send yourself. A deployed strategy locks its spec, so the
thing you tested is the thing that runs.
Real setups are sequences. "The daily closes above its 20 EMA, and then — within four sessions — the 5-minute chart crosses its previous high." A flat rule tree has nowhere to remember which bar the first thing happened on. A flow does.
ALL (every branch must have arrived), ANY (one is
enough), or QUORUM (at least K of the steps before this one). It
sits on the receiving step on purpose — the AND is a statement about what must
have arrived here, not about what the earlier step emitted.
There is no shipped library of "classic patterns" here, and that is deliberate. Every desk has its own definition of a hammer. You write yours, precisely, and it means exactly that everywhere it is used.
/patterns is the library, /patterns/{id} opens one in
the builder, /patterns/new starts an empty one — so the back button
leaves the builder, not the app.
Run any saved strategy over history, then read a report built to stop you talking yourself into a bad idea. The headline is net P&L after costs and max drawdown — not the gross figure that flatters every intraday system ever written.
[0] conditions are evaluated as they form,
and an intraday result is not quietly optimistic.
A card per sector index, leading with that index's own move and listing the constituent stocks behind it. Money rotating out of IT and into metals is a picture, not a table you have to assemble.
A blotter showing positions, P&L and a kill switch looks identical whether the money is real or not — and that is the one thing a trader must never be wrong about. So the mode is in the URL, it is the first control on the page, and each mode carries its own colour.
/live-trading/paper and /live-trading/live are
different addresses. You cannot land on the wrong one by restoring a tab, and a
bookmark always means what it meant yesterday.
Your account, your execution connection, your session control. Market data is already configured for you — there is nothing to buy, plug in or keep alive.
Open the strategy builder and write your entry as conditions — indicators, price sources, crossovers, bar offsets. Set the stop, the target and the trailing stop. Pick the universe. No syntax, no scripts, nothing to debug.
Run it over years of real NSE and BSE bars. Two hundred symbols come back in seconds, with net P&L after every charge, max drawdown, an equity curve and every single trade listed.
Change one thing, run it again, compare. Because a run stores its own parameters, you are always comparing like with like — and because it takes seconds, you actually do the iteration instead of guessing.
Run it forward on live bars with simulated fills and watch it behave. Connect your broker account when you are ready to put the same rules in front of real capital.
No Python layer, no notebook, no queue. Native code with a per-core worker pool sitting directly on columnar bar storage. These are measured numbers from a 16-core machine, not projections.
| Workload | Before | Optimised | Gain |
|---|---|---|---|
| 20 symbols × 2,500 daily bars | 122 ms | 54 ms | 2.2× |
| EMA(20), forming | 295.5 µs | 14.9 µs | 19.8× |
| Supertrend(10,3), forming | 330.1 µs | 33.5 µs | 9.9× |
| ATR(14), forming | 230.3 µs | 37.0 µs | 6.2× |
| ADX(14), forming | 233.9 µs | 42.8 µs | 5.5× |
| RSI(14), forming | 122.4 µs | 23.1 µs | 5.3× |
The fan-out spreads your universe across a worker pool sized to your machine, and deliberately reserves cores for the API, the sockets and the databases — so a heavy backtest never starves the screen you are watching it on.
One-minute and one-day bars in an embedded analytical store, with every other timeframe folded on read. Millions of candles scan without a database server to run, tune or pay for — and a chart switching between two frames that share a base costs no request at all.
Optimisation work is held to a golden trade count and P&L, and every fast path is parity-tested against the plain implementation. A speed-up may never quietly change what a strategy means — otherwise last month's result and this month's are two different questions wearing one answer.
The web client and the Android app read the same data through the same API, and the numbers are held to agreement on purpose — a trader who opens both sees one product.
The full platform: charts, strategies, flows, patterns, backtests, sector scope, watchlists and live trading. Real URLs for every screen, so everything is bookmarkable and the back button behaves.
A native Android client — not a wrapped web page. Sector Scope, watchlists with live prices, and a full candlestick chart with fourteen timeframes, eleven price styles, crosshair, pan, pinch-zoom and a one-tap chart screenshot to your gallery.
A single native binary serving the API, the market sockets, the bar stores and the backtest engine. No interpreter, no runtime to keep alive, and no second language between your rules and the candles they are measured against.
This is a platform that holds the keys to a trading account. The security model was designed first, not retrofitted after the features.
Memory-hard hashing with a per-password random salt. Hashes are never logged and never serialised anywhere.
Opaque, stored only as a hash, and rotated on every use. Presenting an already-used token is treated as theft and revokes the entire family immediately.
An unknown email still runs a full verification against a decoy hash and returns a byte-identical response, so nobody can discover who has an account by timing the login form.
Separate budgets for failed logins, login requests per address, and registrations — plus a flood ceiling charged before routing, so a new endpoint is metered the day it ships.
The live sockets are opened with single-use tickets that expire in thirty seconds. Your access token never appears in a socket URL, a proxy log or a browser history entry.
On Android, tokens are encrypted with AES-256-GCM in the KeyStore and never cross into JavaScript. A compromised bundle cannot read a credential it was never given.
No. Every rule in GTP Algo is built from dropdowns and number fields — pick an indicator, pick what to compare it to, pick the comparison. A finished condition reads back as a sentence. There is no scripting language to learn and nothing to debug, which also means there is no class of bug where the code does something other than what you meant.
Phase 1 covers NSE and BSE — equities and derivatives, priced in INR, on the IST session (09:15–15:30). Exchange holidays and session times come from a maintained trading calendar, so anything that counts trading days counts them correctly. Crypto is a later phase and is not implemented; we would rather say so than list it on a feature grid.
You register in a minute, and your account then sits in an approval queue until our desk reviews it. That is deliberate: this platform connects to real broker accounts, and access to it is granted rather than self-served. You will see a clear status screen while you wait, and if an account is ever turned down you are told the reason rather than left guessing.
You can connect your broker account today — API credentials, the OAuth authorisation round-trip and selecting an active account all work, and at most one broker is active at a time.
Automated order routing is the piece still in build. Everything up to it is live: strategies, flows, backtests, and paper trading that runs your rules forward on live bars with the same fill model the backtest used. We would rather be straight with you here than have you discover it during a session.
They are built to be. Every fill is charged brokerage, STT, exchange transaction charges, GST, stamp duty and your configured slippage, and the report leads with net P&L and max drawdown rather than the flattering gross number. Where one-minute data exists, the engine walks each bar minute by minute so stops and targets fire where they would really have been hit.
What no backtest can model is liquidity on the day, a gap through your stop, or a market that stops behaving the way history did. A good backtest tells you an idea is worth risking money on — never that it is safe.
Traders who already have ideas and want to know whether they hold up. If you have ever said "this setup works, I just can't prove it", this is the tool that settles it. It suits individual systematic traders, small trading teams, and analysts who want sector rotation and screening without building a data pipeline first.
Create your account and we will confirm your plan when we approve it. There is no card required to register, and nothing is charged while your account is in the queue.
No. Market data comes with the platform — years of NSE and BSE history plus the live feed, already synced and ready, with nothing for you to buy or configure. Your own broker account is separate, and you connect that yourself when you are ready to trade through it.
Build the strategy you have been arguing with yourself about, run it over five years of NSE and BSE history, and read the number that comes back after every charge. It takes an evening — and it will either save you a year of tuition or hand you a system you can actually trust.
Free to register · No card required · Approved by our desk, usually within a business day