Configuration

Customize trading parameters and risk management.

Trading Parameters

Edit agent-config.json (created on first run) or use the dashboard:

Setting	Default	Description
max_positions	3	Maximum stocks to hold at once
max_position_value	2000	Maximum $ per position
take_profit_pct	8	Auto-sell at this % profit
stop_loss_pct	4	Auto-sell at this % loss
min_sentiment_score	0.3	Minimum bullish sentiment to consider
min_analyst_confidence	0.6	Minimum LLM confidence to execute trade
min_volume	10	Minimum message volume to consider
position_size_pct_of_cash	20	Max % of cash per position
llm_model	gpt-4o-mini	OpenAI model for analysis

Timing Parameters

Setting	Default	Description
data_poll_interval_ms	60000	How often to fetch signals (ms)
analyst_interval_ms	120000	How often to run trading logic (ms)

Environment Variables

Set these in .dev.vars:

Variable	Required	Description
ALPACA_API_KEY	Yes	Your Alpaca API key
ALPACA_API_SECRET	Yes	Your Alpaca API secret
ALPACA_PAPER	Yes	true for paper, false for live
OPENAI_API_KEY	Yes	Your OpenAI API key
KILL_SWITCH_SECRET	Yes	Secret for HMAC signing tokens

Safety Features

These are enforced by the MCP server's policy engine:

Feature	Description
Kill Switch	Emergency halt for all trading
Position Limits	Max positions and $ per position
Daily Loss Limit	Stops trading after 2% daily loss
Cooldown Period	30-minute pause after losses
Approval Tokens	Orders expire after 5 minutes
LLM Confidence Gate	Trades require minimum confidence
No Margin	Cash-only trading
No Shorting	Long positions only

Tuning Tips

Agent not making trades?

• Lower min_analyst_confidence to 0.5
• Lower min_sentiment_score to 0.2
• Check max_positions — might be at limit
• Check research results in dashboard

Too many trades?

• Raise min_analyst_confidence to 0.7+
• Raise min_sentiment_score to 0.4+
• Lower max_positions

High OpenAI costs?

• Increase data_poll_interval_ms
• Use gpt-4o-mini (default, cheapest)