TT Assistant · connect in one click
Pick your AI, click once, sign in — Tradetron's tools appear in your chat. No install, no code.
Opens Claude with the server pre-filled. Sign in with Tradetron, approve — done.
https://mcp.tradetron.tech/mcpTurn on Developer mode (Settings › Security and log-in), then Plugins › + and paste the URL. Needs a paid plan.
https://mcp.tradetron.tech/mcpCopy the command, paste in your terminal, then type /mcp to sign in.
claude mcp add --transport http tradetron https://mcp.tradetron.tech/mcpOpens Cursor and configures the server. Click Connect and sign in.
https://mcp.tradetron.tech/mcpGemini, VS Code, Antigravity, Grok and more — add the URL as a remote MCP server.
https://mcp.tradetron.tech/mcpPaste one URL into Claude, ChatGPT, Gemini, Cursor or Codex, sign in with your Tradetron account, and Tradetron's tools appear in the chat. Describe a strategy in plain English and it gets built in your account. Backtest it. Change one leg of it. Ask why it didn't trade this morning. Tradetron charges nothing for the connector — it runs on the AI subscription you already pay for.
strategies built by traders through the Tradetron MCP
MCP — the Model Context Protocol — is the open standard AI apps use to reach outside services. Tradetron publishes one. You add its address to your AI once, sign in on Tradetron's own page, and the tools are in every chat after that. Nothing to download, no key to copy, no code to write.
There is no new interface to learn and no second place to log in. The chat window you already have open gets your Tradetron account attached to it.
No add-on, no plan upgrade. The thinking is done by your own Claude, ChatGPT or Gemini subscription, so a conversation costs exactly what that subscription already costs you. On our side, a Tradetron account is the whole requirement.
It builds, tests, edits and explains strategies, and can deploy them in Live Offline, which runs on live prices without sending a single order. Placing real orders stays with you, inside Tradetron. Your money never leaves your own broker.
Six jobs you would otherwise click through the wizard, the backtest queue, the builder canvas and the notification log for. The line in quotes is what you type.
“Short NIFTY ATM straddle, 50-point wings, exit at 30% profit or a 3% MTM stop.”
You get a real Advanced Strategy Builder template in your account, with a link to open it. Every keyword is resolved and every field validated before anything is saved.
“Backtest it from January 2022.”
It runs on Tradetron's own engine and hands back the full report link. It is priced like any Tradetron backtest — ₹20 a run, whatever the date range — and you see the price and confirm before anything is charged.
“Add a hedge leg to set 2 — don't rebuild it.”
Surgical edits are round-trip verified. If the platform would silently drop something, the edit fails loudly instead of quietly leaving your strategy half-changed.
“My straddle is in error and didn't trade this morning. What happened?”
It reads the engine log and the broker log together and names the layer at fault — your logic, or the broker's rejection. That is usually the difference between an hour of guessing and a one-line fix.
“Rank my strategies, worst to best.”
P&L, win-days, payoff ratio and drawdown for one strategy or the whole book. Descriptive reads of your own history — never a recommendation to buy, sell or hold.
“Deploy it and let me watch it for a week.”
It deploys in Live Offline: the strategy runs on live prices through your connected broker and you see every signal, but no order is sent. Ownership-checked, and the connector cannot switch a deployment to placing orders.
Those are the shapes, not the list. Every tool, grouped by job and with real responses, is in the tool reference.
Test a trading idea in one sentence. There is no strategy to build first: the assistant turns the idea into a strategy, replays it on past prices, and tells you how it would have done.
“What would have happened if I bought RELIANCE whenever its RSI dropped below 30 and sold it a week later?”
Name the stock, when to buy, and when to sell: after a number of days, when another rule turns true, or both. Rules can use RSI, moving averages and price on daily candles. If you leave out when to sell, the assistant asks.
Covers NSE stocks and ETFs on daily candles, buying only. Holding periods are
calendar days. The preview uses the same price data as the engine, but the
engine's report is the authoritative result. Your assistant uses the
tt_backtest_idea tool for this; you do not need to name it.
No hunting for keyword names you don't remember. Tell it what you want; it picks the blocks, names the legs, sets the conditions, and shows you exactly what it built before anything is saved.
When a description is ambiguous — “stop loss on the short call” could mean on premium or on the underlying — it asks one question instead of guessing. That is the whole game: nothing gets dropped quietly.
It resolves more than 8,600 tradable underlyings by name, so “Bank Nifty weekly”, “Sensex”, “Crude” and “Reliance” all land on the right instrument without you looking up a symbol.
Short NIFTY ATM straddle, Thursday 9:30, 50-point wings, 30% profit target or 3% MTM stop.
Built — 4 legs (2 short ATM, 2 long 50-point wings), entry 09:30 Thursday, target 30% on net premium, stop 3% MTM. One question: is the 50-point hedge a strike distance or a premium? I've assumed strike distance.
Strike distance is right. Backtest it from April 2024.
That's one run on Tradetron's engine: ₹20, whatever the period. Go ahead?
Yes.
Queued. I'll post the report link when it finishes — the strategy is saved in your account either way.
One address, added to the AI app you already use. Sign-in is OAuth: you log in on Tradetron's own page, and your password never reaches the AI app.
One command, then /mcp in a session to sign in.
claude mcp add --transport http tradetron https://mcp.tradetron.tech/mcpAdd the server to ~/.codex/config.toml. The browser opens once for sign-in.
[mcp_servers.tradetron]
command = "npx"
args = ["-y", "mcp-remote", "https://mcp.tradetron.tech/mcp"]Add a remote MCP server pointing at the same URL. In Cursor that is
Settings → Tools & MCP → New MCP server (or the one-click
button in the connect window); in VS Code it is a .vscode/mcp.json entry.
{ "servers": { "tradetron": { "type": "http", "url": "https://mcp.tradetron.tech/mcp" } } }Point it at the same URL as a remote HTTP MCP server. If a client only supports local servers, bridge it — this works everywhere:
npx -y mcp-remote https://mcp.tradetron.tech/mcpEvery tool is also a plain Python method that returns a dict — useful for scripts, notebooks and cron jobs. Sign-in is the same OAuth flow, cached after the first run.
pip install ttmcp
from ttmcp import TradetronMCP
tt = TradetronMCP() # opens your browser, once
tt.tt_portfolio_report(from_date="2026-07-01", to_date="2026-07-31")Menu names move between app versions (ChatGPT calls a connector a Plugin); if you can't find Connectors, search your AI app's settings for MCP.
An algorithmic trading platform used across Indian markets — NSE equities and F&O, plus MCX. You build a strategy once, backtest it against historical data, and let it run against your own broker account. A Tradetron account is free to open, and it is all you need to connect.
Create a free accountThe connector is a thin, checked client on top of Tradetron. It cannot touch your account in ways you wouldn't.
Every spec is checked against the locked schema first. Wrong types, unknown keywords and dropped legs are caught before anything reaches your account.
Before reading any strategy's logs or P&L it proves the strategy is yours. If it can't confirm ownership — or the check itself fails — it refuses and reads nothing.
Structural edits are round-trip verified. If the platform would silently drop content, the edit is rejected rather than left half-applied.
It can deploy a strategy in Live Offline, which sends no orders. It cannot place, change or cancel an order, move funds, or change how a running strategy trades.
The assistant never places an order. It can create and edit strategies, backtest them, and deploy them in Live Offline, which runs on live prices without sending orders. Trading real money stays a deliberate action you take yourself, signed in to Tradetron. Nothing here is investment advice.
Tradetron charges nothing for the connector. There is no add-on and no plan upgrade — the AI's thinking is billed by your own Claude, ChatGPT or Gemini subscription, exactly as it is when you use that app for anything else. The only thing you pay Tradetron for in this flow is a backtest, because it runs on Tradetron's engine: ₹20 a run, whatever the date range, the same as on the Backtest page.
No. A Tradetron account is enough to connect and to build, validate and save strategies. Your plan still governs what you can do with those strategies afterwards — how many you can deploy, and so on — exactly as it does when you build them by hand.
No — that's the point. You describe the strategy in your own words and the connector handles the keyword lookups, the schema and the field-by-field wiring. What you end up with is an ordinary template in the Advanced Strategy Builder that you can open, review, tweak by hand and deploy. You just didn't have to assemble it.
Whichever one you already pay for. The tools and their behaviour are identical underneath — the AI app is only the conversation layer. Claude, ChatGPT, Gemini, Cursor, Codex, Claude Code and other MCP clients all connect to the same server at the same URL.
No. Its job is the strategy lifecycle — create, validate, backtest, edit, diagnose, report, and deploy in Live Offline, which runs a strategy on live prices without sending orders. It cannot place an order, move funds, or pause, resume or square off something that is already running. Those stay with you in Tradetron. Funds are held in your own brokerage account throughout; Tradetron never custodies money.
Yes, once, when you add the connector. Tradetron's own sign-in page opens, you log in and approve access, and your AI app never sees your password. After that it acts as you in every chat. To cut access off, remove the connector in your AI app.
Once you sign in, it acts as you and only as you: your strategies, your deployments, your logs, your reports. Every read is ownership-checked server-side, so it cannot reach another user's data even if asked.
Not today. The engine log records which condition fired, not the number that made it fire — indicator values are computed live on each tick to answer true or false, and aren't stored for retrieval afterwards. Capturing the values behind each decision is on the roadmap.
It won't guess. Broker, instrument and pricing answers come from authoritative catalogs rather than the model's training data, and product answers carry citations. When something genuinely needs a person, it can open a real support conversation and relay the replies back into the chat — only after you say yes.
Paste one URL into the AI you already use, and build your first Tradetron strategy tonight.
Build, backtest & deploy strategies in plain English — no code.
Paste into Customize › Connectors › Add custom connector, then Connect. Works on every Claude plan.
Claude says it already exists? You (or your team) added it before — open Customize › Connectors and use that one.
Settings › Security and log-in → turn on Developer mode → Plugins › +, paste URL, auth OAuth, Create. Then pick Tradetron from + in the message box. Requires Plus / Pro / Business / Enterprise — not Free.
Run it, then start claude and type /mcp to sign in.
Add to ~/.cursor/mcp.json, then Connect under Tools & MCP.
Standard remote MCP (HTTP + OAuth). Verified with Windsurf, VS Code, Cline, Gemini CLI. Local-only clients bridge via npx mcp-remote.