- fintech
- FinTech & Algorithmic Trading
TraderOps: The Trader's Operating System — Backtest, Paper & Live on One Engine
Built a multi-broker trading platform with paisa-exact backtest-to-live parity, a no-code strategy builder, and Telegram + MCP control across eight Indian brokers.
- FinTech Engineering
- Trading Infrastructure
- AI & Automation
At a glance
- Sector
- FinTech & Algorithmic Trading
- Type
- fintech
- Services
- FinTech Engineering · Trading Infrastructure · AI & Automation
The problem
What it had to solve
Retail and semi-professional options traders juggle a broker terminal for execution, a separate backtesting tool that never matches live fills, and manual spreadsheets for risk. Strategies that look profitable in a backtest fall apart live because the trigger math, slippage, and order-type behaviour differ. Managing stop-losses across multiple broker accounts by hand is error-prone, and there is no safe kill switch when a day goes wrong.
Disciplined, salaried traders want automation they can trust without writing code or babysitting a screen. The goal was one engine where a strategy is authored once and behaves identically in paper and live — so what you test is exactly what you trade — while remaining broker-agnostic and safe by construction.
What we built
How it works
TraderOps is a Python/FastAPI engine over PostgreSQL with a broker-abstraction layer that normalises order placement, modification, and reconciliation across every supported broker. SL/target/trailing trigger prices are anchored to the strategy's intent so the trigger ladder is bit-exact across live, paper, and backtest by construction — slippage moves the fill, never the trigger.
A single-orderbook reconcile loop is the only path that updates trade state from the broker, eliminating race conditions between per-order polling and event handling. Market data flows from one shared platform feed rather than per-user tickers, and execution is decoupled from the data source so any broker can execute against a common feed. Every action — Telegram command, MCP call, webhook, or scheduler — lands in the same audited trade pipeline.
Calls we would still defend
Paisa-exact parity — trigger prices are anchored to expected entry so backtest, paper, and live produce an identical trigger ladder — the core trust guarantee of the platform.
Broker-agnostic layer — a frozen capabilities model per broker (tag format, product/exchange maps, order semantics) means engine code never branches on broker name, so adding a broker is additive and safe.
Control anywhere — a read/write MCP server and a Telegram bot let traders monitor and execute strategies conversationally, with a two-step confirm gate on every order-placing action.
Safe by construction — mode-scoped kill switches, per-strategy capital limits, and a scheduled MIS auto-square-off protect capital without manual intervention.
Built with
Python + FastAPI
Powers the async execution engine, scheduler, and REST API that every client (web, Telegram, MCP, webhooks) shares.
PostgreSQL + SQLAlchemy
Durable, transactional store for trades, strategy runs, and audit events — the single source of truth reconciled against the broker.
Model Context Protocol (MCP)
Exposes read + write trading tools to AI clients so strategies can be authored, fired, and monitored conversationally behind a confirm gate.
Telegram Bot API
Lets traders run, monitor, and exit strategies from chat — no screen required — with rich fill and P&L notifications.
Where it landed
What the build changed
8
Broker integrations
Indian brokers behind one execution API; a strategy is authored once and runs on any
Paisa-exact
Backtest-to-live parity
backtest, paper and live share one trigger ladder anchored to the expected entry price
3
Execution modes
paper, live and backtest, switched without rewriting the strategy
FAQ
The questions this build raises
What was actually built, the constraints it had to meet, and how it holds up in use.
SL, target, and trailing trigger prices are anchored to the strategy's intended entry rather than the actual fill, so the trigger ladder is computed identically in backtest, paper, and live. Slippage moves only the fill price — never the trigger — which keeps results paisa-exact across modes.
Eight right now — Zerodha, AliceBlue, Tradejini, Zebu, Dhan, Delta, Kotak Neo and Motilal Oswal. All eight are Indian brokers, and a broker-agnostic layer normalises how each one behaves, so the same strategy runs on any of them.
Yes. The strategy builder composes ORB (Opening Range Breakout), WAT (Wait-And-Trigger), re-entry, and multi-leg option strategies through configuration — no code required — and you can control everything from Telegram or an MCP-connected AI client.