AI Agent for Prop Firm Challenges · Setup Guide
Trade with an AI agent for prop firm challenges.
Create a free Propr account, then paste the setup prompt into Claude or ChatGPT. Your AI guides you through connecting Tally's MCP and setting up the hosted agent.
- 1. Create your trading accountCreate free Propr account (opens in a new tab)
Sign in with Google, choose Free Trial, then return here for step 2.
- 2. Set up your AI agentGet the setup prompt
Already have an account? Start here. Copy the prompt and give it to your AI.
I earn a commission if you later buy a challenge through this link, at no extra cost to you.
Trading loses money for most people, and most people who buy a prop challenge do not pass it. This setup guide is not advice, and nothing here promises a result. Run it on the free trial first and never buy a challenge with money you need.
The setup prompt
Paste this into a normal Claude or ChatGPT chat. It starts with Tally's AI connector, then guides you through the account and Propr trial setup. Your AI helps you follow the steps; Tally runs the agent.
Help me set up Tally's hosted Propr trading agent and connect it to this AI app. Guide me one step at a time. Start with the Tally connector before asking me to create a Propr account. Use Tally's current instructions at https://tally.markets/connect?utm_source=prosperitylabs&utm_medium=agent-guide&utm_campaign=tally-hosted-v1 and its MCP URL https://tally.markets/api/mcp.
First, check whether Tally's tools are already available here. If not, guide me through adding the connector in this AI app. In ChatGPT, use its app directory; in Claude, show me how to add a custom connector named Tally with the MCP URL and leave OAuth fields empty. If you cannot add it yourself, tell me the exact steps to do it. Ask me to create or sign in to my Tally account and approve its connection. Do not say the connector works until you can call get_started. If this AI session cannot use the connector, explain the supported way to continue without pretending it is connected.
Next, guide me through Tally's dashboard setup. If I need a Propr account, show me https://app.propr.xyz/r/AITRADINGTOM and help me choose a free trial. Show me how to create a Propr API key in Propr Settings > Developer. I will enter the key in Tally's dashboard, not this chat. Help me select the exact trial account and markets, and review Tally's strategy, risk limits, and Sentinel subscription ($4.99/month or $49/year; Propr challenge fees are separate).
Ask me to review the account and settings before I decide whether to subscribe and explicitly authorize trading in Tally. Subscription alone does not start trading. Tally hosts the agent and its hourly risk checks; neither this chat nor the MCP connector places orders. Do not claim the agent is running until Tally confirms it. After the subscription is active, call get_checkup and tell me the real agent status, positions, protective stops, risk-limit headroom, and anything needing attention. If a step is blocked, tell me the exact next action. Do not install a repository or create a cron job.When it asks you to sign in or approve the connector, follow the steps in your AI app. Tally's Connect your AI guide shows the same flow. If you create a Propr account through the link in the prompt, I may earn a commission if you later buy a challenge, at no extra cost to you.
What the prompt helps you do
Connect Tally to your AI. Your AI checks for Tally's MCP connector and shows you how to add it if needed. Tally asks you to sign up or sign in and approve the connection. You can then ask your AI for a checkup; the connector itself cannot place trades.
Set up your Propr trial. In Tally's dashboard, enter your Propr API key, then choose the exact free-trial account and markets. Your key goes into Tally's setup form, not your AI chat.
Review and authorize. Tally shows your selected account, strategy and risk limits. Sentinel costs $4.99 per month or $49 per year; Propr challenge fees are separate. Paying for Tally does not start trading. You explicitly authorize the agent on your chosen account.
Tally's service runs the agent and hourly checks. Your dashboard is where you start or stop it; your AI can read its status and journal once the account is subscribed and connected.
Rules in the DIY kit
These are the standalone kit's rules for people building their own agent. Tally shows its current strategy and risk limits in its dashboard before you authorize it; review those there.
The system
Daily trend following across a chosen set of Propr markets, including crypto, equities, indices, commodities and FX. The agent trades long and short. It uses a Turtle-inspired 20-day breakout and 10-day exit, adapted for daily closes, no adding to positions and the challenge limits below. It is not the original Turtle system.
The challenge it is built for
Propr's Bronze Classic one-step, a 25,000 USDC account for a $275 fee. Ten percent profit target, three percent daily loss limit measured from the start-of-day balance, six percent maximum drawdown measured from the starting balance and never moving. Both loss limits are checked against equity, including open P&L. No time limit, no minimum trading days. The risk settings below use those limits. If you run a different challenge, update the DIY configuration first.
Markets
Choose a watchlist during setup from the markets available to your Propr account. Each market must support API orders and protective stops, with enough reliable daily history to calculate and test the rules. Use the exact market identifiers, including any namespace prefixes. Fix the list before a paid challenge. A familiar ticker alone does not confirm that an instrument is supported.
Timeframe
Completed daily bars matched to each Propr market's underlying Hyperliquid contract. Check each instrument's trading hours and data freshness before placing orders. The kit runs one tick each hour and scans for new daily signals once a day at or after 00:10 UTC. Each tick checks equity, reconciles orders and confirms protective stops. Between ticks, confirmed reduce-only stops resting at Propr are the only automated protection.
Entry
Go long when the daily close is above the highest high of the previous 20 days. Go short when the daily close is below the lowest low of the previous 20 days. One position per market, no adding.
Initial stop
Two times the 20-day ATR from the actual fill price, placed on Propr as a reduce-only stop-market order as each fill arrives. ATR uses Wilder smoothing: start with the average of 20 daily true ranges, then use (19 times the previous ATR plus today's true range) divided by 20. Use the completed signal day's ATR. Round stops toward the fill price to the allowed tick; never widen them.
Position size
Risk 0.4 percent of the starting balance per trade, which is $100 on the 25,000 account. Size equals that dollar risk divided by two times ATR, in the instrument's order units, adjusted for its contract multiplier and rounded down to its allowed size. Hold at most three positions at once. Their planned stop risk totals 1.2 percent of the starting balance, before fees and slippage. That is not a guarantee against a daily breach. Propr applies leverage automatically. The agent caps total gross position value, including pending entries, at two times current equity by reducing or skipping entries. It never changes leverage in account settings. Most positions will use far less.
Exit
Close a long when the daily close is below the lowest low of the previous 10 days. Close a short when the daily close is above the highest high of the previous 10 days. Close earlier if the stop fires. There is no profit target for an individual trade.
Daily halt
Propr's daily loss limit is checked against equity, including open P&L, using the balance at 00:00 UTC as the day's reference. At the next hourly tick after equity is observed two percent below that start-of-day balance, the agent cancels pending entries and opens nothing more until the next daily run after the reset. Existing stops and exits still apply. The challenge fails if equity touches three percent below the start-of-day balance at any point. Halting new entries leaves a one percent buffer, but open positions can still breach it.
Kill switch
Propr's maximum drawdown counts open positions, so it is measured on equity. At the next hourly tick after equity, including unrealised losses, is observed 4.5 percent below the starting balance, the agent blocks entries, cancels pending entries and closes every position. It keeps protective stops until positions are confirmed flat, then cancels remaining orders. It only reports shutdown complete after verifying both. Failed or uncertain requests keep the halt active; the next tick reconciles and retries. It writes a halt report. The challenge fails at six percent, and touching either limit closes the account for good. The agent does not restart itself.
Orders
The default entry is an immediate-or-cancel limit order at the current ask for a buy or bid for a sell. It attempts the trade immediately; any unfilled remainder is canceled. Make one attempt per market per signal day, with no chasing or topping up partial fills. Market entries are an explicit setup option, not a claim of better performance. Use reduce-only market orders for exits and emergency closes. Protect every fill, including partial fills, with a reduce-only stop-market order. Respect each market's trading hours, price increments and size limits. Fees, funding, spreads and slippage vary by instrument and must be included in testing.
Free trial first
Test the scheduled agent on Propr's free trial account for thirty days before considering a paid challenge. It submits orders through Propr's API, manages positions and places stops. These are trades in Propr's simulated account. If no entry rule triggers, it waits. The daily report records what it did.
How it behaves
The agent enters after a price breakout and exits on a reversal or stop. Sideways markets can produce repeated losses. Sustained trends give it a chance to hold a winning trade. Propr has no time limit, so there is no need to increase risk to meet a deadline.
Build your own agent (optional)
Want to run the standalone open-source kit instead of Tally? This is the more hands-on path. It needs a cloud coding environment, private credential storage, persistent state and a verified schedule you control. The kit and Propr trial are free; AI app access depends on your plan.
Open a Propr free trial. Sign in with Google and choose Free Trial. In Propr Settings, open Developer and create an API key when your coding agent asks. Keep it private.
Start a free Propr trial (opens in a new tab)I earn a commission if you later buy a challenge through this link, at no extra cost to you.
Use the DIY cloud setup guide. Give it to Claude Code or Codex in a cloud coding task. The pinned beta.13 kit installs the shared framework, runs the tests and checks its environment. It asks which Propr markets you want to trade and uses the scheduled `propfirm tick` command.
Connect your trial account. After the offline tests pass, your agent shows you where to create a Propr API key and how to add it as PROPR_API_KEY in your private cloud environment settings. The coding agent can use that key; keep the environment private. Your agent retrieves your Propr accounts and asks you to choose which one to use; you do not need to find an account ID. It checks your account and runs the backtest. Your key, settings and reports stay in your private setup, outside the public kit repository.
Review the checks and start trading. Once the account and scheduler checks pass, your agent runs a read-only check, then one supervised trial tick. The verified schedule runs future ticks hourly and reports its last result. It sends an order only when the rules signal, protects fills with stops and checks account risk on each tick. Check the scheduled runs and your orders in Propr. No signal means no new trade.
This is a trial-testing beta with order execution built in. A real Propr fill, protective stop and continued operation from a viewer-owned schedule still need to be verified together. Your agent should report any missing capability or account field before starting. Read the daily reports for thirty days before considering a paid challenge.
Setup details: Cloud setup guide.
DIY kit: moving to a challenge
After thirty days on the free trial, review the reports and the backtest's challenge failures. If you decide to pay for a challenge, these rules are set for the Bronze Classic one-step.
Have your agent change the account ID only when you explicitly choose to use a paid challenge. Check that the total exposure cap is two times equity. Run one scan by hand and read the report before enabling the schedule. Confirm that the scheduled tick is running, its reports are current, and Propr still holds the protective stops between ticks.
I earn a commission when you buy a Propr challenge through this link. It costs you nothing and it is how this kit stays free. Propr did not write or review the rules. Propr's accounts, including funded ones, are simulated. The payouts are real USDC.
What to expect
The agent can lose money and fail a challenge. A backtest shows how the rules behaved on historical data. It does not establish a win rate or tell you when the agent will reach the profit target.
Use the free trial to check that orders, stops and risk controls work as intended. Read the daily reports, including the losing trades. Thirty days is a starting point for testing, not proof that the agent will pass a paid challenge.
Tally hosts the primary setup shown above. Its account connection and AI checkup should be confirmed in your own account before you rely on them. The optional DIY cloud flow still needs a viewer-owned Propr trial order and stop acceptance test; it is not a verified hands-off installation.
Build the agent with me in a weekend.
Join me live on Saturday, October 24, and Sunday, October 25, from 11am to 1pm Eastern each day.
We'll use Claude Code and Codex cloud sessions to build and test the kit, then connect a Propr free trial with your own API key. We'll check the hourly scheduled run, private state and protective orders before any trial trading.
You get the recordings, the finished code repository, and access to a private group for the first thirty days of your own challenge.
The course costs $197. It's a small group. If you're the only person who books, we'll work one on one. You can get a full refund any time before the first session.
Book a seat · $197 (opens in a new tab)The DIY kit is free. The course gives you time to build it with me and ask questions.
