Describe the bot in plain English

How this works

This demo runs a rule-based parser entirely in your browser — it reads markets, barriers, stakes, ladders, virtual-trade gates and limits straight out of your sentence, then fills the Guided tab so you can see and adjust everything it decided. Nothing is sent anywhere.

In the live version this same box calls a language model, which handles free-form descriptions the rules miss (“something safe for a $20 account that trades a few times an hour”) and can invent combinations rather than matching keywords. The output is the same strategy spec, so the compiler below never changes.

Upload a bot to repair, rebrand and harden

Your strategy logic is not rewritten. The upgrader repairs what is structurally broken, replaces another author's branding with yours, and wraps the result handling in a safety stack — a stake ceiling, a loss-streak reset and a take profit / stop loss that genuinely stop the bot.

Safety settings to inject

The ceiling is applied to the bot's own stake variable after its sizing has run, so whatever its maths produces is bounded. The streak limit resets that variable to its starting value after this many losses in a row.

Audit

What it can and cannot do

It reliably fixes structure: a wrong root namespace, duplicate block ids, orphaned and disabled blocks, incomplete purchase blocks, an out-of-date market block, and variables declared but never used. It rewrites branding, and it can find the stake variable in most bots by reading the amount socket on the trade options.

It will not invent strategy. If a bot reads a variable nothing ever assigns, or gates a branch on something meaningless, the audit says so and leaves it alone — only the author knows what was intended. Read the report before you trust the output, and test on a demo account first.