No-Code Agent Manual

Overview

The no-code agent builder is a 6-step wizard that transforms your investment thesis into a live trading agent. You describe your strategy in plain English, upload supporting research, configure risk parameters, choose an AI model and analysis tools, then deploy to paper trading. The platform's LLM interprets your instructions on each trading tick and makes buy/sell/hold decisions autonomously.

Step 1: Writing Instructions

Your instructions are the core of your agent. Think of them as briefing an analyst — you describe what to look for, how to evaluate opportunities, and when to act. The LLM reads these instructions on every trading tick and uses them to make decisions.

I want to invest in US large-cap technology companies
that are undervalued relative to their free cash flow
growth. Rebalance monthly, hold 15-25 positions, and
never let any single position exceed 8% of the portfolio.
If the portfolio draws down more than 12%, reduce
exposure by 50%.

Focus on mid-cap US equities showing strong price
momentum over the last 3 months. Use RSI to avoid
overbought names (RSI > 70). Rebalance weekly, hold
10-15 positions, equal-weight. Max drawdown tolerance
is 15%.

Build a quality-focused portfolio of S&P 500 stocks.
Prioritize companies with high Piotroski F-Score (>= 7),
strong free cash flow yield, and low debt-to-equity.
Use analyst consensus as a secondary signal. Rebalance
monthly, hold 20-30 positions, conviction-weighted.

Step 2: Uploading Knowledge

Knowledge documents ground your agent in your domain expertise. Upload research reports, financial models, screening criteria, or any data that should inform your agent's decisions. Documents are processed and made available to the LLM as context during each trading tick.

Step 3: Configuration

Configure the parameters that define your agent's trading universe, position sizing, rebalancing frequency, risk limits, and starting capital.

Step 4: Model Selection

Choose the foundation LLM that powers your agent's decision-making. The model receives your instructions, knowledge documents, market data, portfolio state, and tool outputs on each tick, then produces a trading decision.

Step 5: Tool Selection

Tools give your agent access to real-time analysis capabilities. The LLM decides which tools to call based on your instructions and the current market context. All tools use real market data from Alpaca and Financial Modeling Prep (FMP).

Step 6: Review & Deploy

The final step shows an AI-generated summary of your agent configuration, including risk warnings and potential issues. Review everything carefully before deploying.

How Agent Execution Works

Once deployed, your agent runs on a scheduled tick cycle. On each tick, the platform:

1. Fetches current market data for all symbols in your universe (prices, daily changes)
2. Reads your agent's current portfolio state (cash, positions, unrealized P&L)
3. Assembles a prompt with your instructions, knowledge documents, market data, portfolio state, and available tools
4. Sends the prompt to the LLM, which can call tools (RSI, DCF, etc.) and reason about the data
5. The LLM returns target portfolio weights or individual buy/sell decisions
6. The constraint engine validates the decisions against your risk limits (max position size, sector concentration, max drawdown)
7. The trade deriver converts target weights into actual orders, accounting for current positions
8. Orders are submitted to the broker (Alpaca) for execution

Platform Limits