The Tradetron MCP is the connector that turns ChatGPT, Gemini, Grok or Claude into a working quant desk on your account. This page is the power-user map: every tool, grouped by job, with worked examples of what really comes back — and an exact account of what the debug logs can and cannot show. New here? Start with the friendlier TT Assistant page.
Grounded answers about the platform, no account needed — plans, brokers, instruments, market hours.
Describe it; the MCP resolves keywords, validates the spec, and creates a real template in your account.
Where power users live: diagnostics that fuse engine and broker logs, surgical edits, portfolio-level review.
You never call these by name — you just ask, and the assistant picks the right ones. The names are here so you know exactly what’s under the hood.
Pulls a worked pattern, resolves every keyword, emits the exact template, creates it in your account with a link to open in the builder. Large multi-set books handled too.
Every spec validated against the locked schema before saving — wrong types, unknown keywords, silently-dropped legs caught locally. Custom Python is security-checked.
Run a backtest from the chat and get the full FastBT report link back — 6.5 years of audited data, ₹20 flat. Sweep parameters, poll long runs, check backtestability up front.
Fetch, edit scalar fields, replace whole sets, or apply a surgical patch to one leg or condition. Structural edits are round-trip-verified — silent drops fail loud. Export to an importable file.
Deploy any strategy to paper trading — never live, that stays with you — and let it run against real market data. Ownership-checked, so it only ever touches your own account.
One question fuses the deployment record, engine notification log and broker log into a diagnosis that names the layer at fault. Drill into either log, or the run counters, when you need to.
P&L for one strategy — win-days, payoff ratio, max drawdown — or the whole book ranked winner-to-loser. Formatted reports and exports. Descriptive reads, never buy/sell advice.
Authoritative, deterministic lookups — not training-data guesses. Broker support, symbol tradability by exchange, step-by-step broker setup, and your watchlists resolved, read and modified.
Grounded, cited answers to any Tradetron question; market open/close for any date; your profile, plan and marketplace subscriptions. When the bot can’t help, it opens a real support conversation and relays the thread — with your say-so.
Three real flows — build, debug, review. The sentence you type, the tools the assistant reaches for, and the shape of the response. Values are illustrative of the format.
{ "ok": true, "data": { "id": 9987412, "name": "NIFTY Weekly Short Straddle", "edit_url": "https://tradetron.tech/strategies/9987412/edit", "dry_run_verified": true, "validation": { "warnings": [] } } }
You get a clickable edit_url — the strategy is already in your account, ready to review in the Advanced Strategy Builder. Nothing was saved until it validated clean.
DIAGNOSIS · SID 30288650 · "NIFTY Weekly Short Straddle" status Error-Execution mode live window last 24h ENGINE LOG (what the strategy decided) 09:20:01 Entry condition met — Time ≥ 09:20 09:20:02 Placing SELL · NIFTY 24800 CE · NIFTY 24800 PE 09:20:04 ✕ order rejected at broker BROKER LOG (what the broker actually did) 09:20:04 Zerodha · order status REJECTED 09:20:04 RMS: "margin exceeds available limit" ANALYSIS Not a strategy-logic fault. Your entry fired correctly; the broker blocked the order — insufficient margin. FIX Add funds or reduce lots, then re-enter. Logic is sound.
Before it reads a single line, the MCP verifies the SID is yours — ownership-checked and fail-closed. The card names which layer broke, so you don’t spend an hour editing a strategy that was never the problem.
DEPLOYED · 6 strategies ● Active NIFTY Weekly Short Straddle live ● Active Bank Nifty Iron Condor live ● Active Gold Calendar Spread live ⏸ Paused Reversal Scalper paper ✕ Error-Exec Momentum Futures live ○ Exited Expiry Day Butterfly live NEXT Just say "why is Momentum Futures in error?" — the engine + broker logs get pulled and the fault named (that's Flow 02, in the same chat).
One question surfaces every deployment and its state — Active, Paused, Blocked, Error-Execution, Exited — with live vs paper mode. Name any strategy and the assistant drills in: diagnose it, read its performance, or manage the deployment.
The debug tools are the most misunderstood part of the connector, so let’s be exact about where the data comes from and where its edges are.
Why: those numbers are evaluated live — every tick, in memory — purely to decide whether a condition is met. The log records the decision, not the arithmetic behind it, so there’s no stored per-candle value to fetch after the fact. Capturing the operand values behind each decision is on the roadmap.
Step-by-step setup for each AI app — with plan requirements — lives on the TT Assistant page ↗. Read-only tools (plans, brokers, docs, drafts) work before you sign in; saving strategies and reading your logs, deployments and P&L need the OAuth connection.
Save this once and stop re-typing the same guardrails. It tells any AI — Claude, ChatGPT, Gemini — how to work on your Tradetron strategies: ask before building, look up every keyword, validate and read back what was stored, and never touch a deployed template without telling you what that means. Drawn from real measured behaviour of the tools on this page.
Prefer a file? Download tt-mcp-instructions.md — same content, save it straight into your project.
Build, backtest, paper trade, debug and review — all from one conversation, in the AI you already use.