Hedge Fund Agent
Most people either pick tokens themselves with no real framework, or pay a fund 2% of their money every year plus 20% of whatever it makes: win or lose on the first fee, lose either way on the second. The Hedge Fund Agent runs a real multi-analyst research process against a strategy you define, at half that cost, on paper first.
Paper trading only. The Hedge Fund Agent tracks simulated positions and P&L. It does not place real trades. This is a research and strategy-testing tool, not live fund management.
What it does
- Builds a strategy from your instruction: pick your own tickers, or let it choose based on a time horizon
- Runs every candidate through 18 named analysts before forming a view
- Tracks simulated positions, take-profit and stop-loss, and re-checks them automatically
- Backtests any strategy against SPY as a benchmark, with Sharpe ratio, Sortino ratio, and max drawdown
- Reports fees exactly as they'd apply to a real account, even though no real money moves
The 18 analysts
Every asset gets scored by three groups of analysts before the agent forms a combined view. Each analyst votes bullish, bearish, or neutral with a confidence score, and the synthesis is a confidence-weighted blend, not a vote of one.
| Group | Count | What they look at |
|---|---|---|
| Quant | 5 | Technicals, fundamentals, valuation, growth, price sentiment |
| Value | 6 | Buffett, Graham, Munger, Pabrai, Fisher, and Damodaran-style value screens |
| Macro | 7 | Druckenmiller, Burry, Wood, Lynch, Ackman, and Taleb-style macro reads, plus news sentiment |
None of this runs through a language model. It's deterministic: the same inputs always produce the same signal. The LLM only writes the plain-language summary of what the analysts already decided.
Try it
1. "Start a strategy: trade SOL and BONK for 3 months, take profit at 20%, stop loss at 10%" → Runs all 18 analysts on both tokens, opens simulated positions, and saves your take-profit and stop-loss rules.
2. "Don't tell me what to pick, just find something for a 6-month horizon" → The agent's own horizon-based picker selects candidates and runs the same analyst process on them.
3. "Backtest this against the last year" → Historical simulation with an SPY benchmark, Sharpe ratio, Sortino ratio, and max drawdown.
4. "How are my positions doing?" → Simulated fills and running P&L for every open position, checked automatically every 4 hours.
Fees
| Management fee | Performance fee | |
|---|---|---|
| Traditional hedge fund | 2% annual | 20% of profit |
| Hedge Fund Agent | 1% annual | 10% of profit, above your high-water mark |
The performance fee only applies to new profit above your previous high point. A strategy that loses money one period and recovers it the next doesn't pay twice.
Limitations
- Paper trading only: no real order is ever placed, so results don't include real slippage or execution risk
- A backtest describes the past, not a guarantee of what the strategy does next
- The horizon-based picker is a starting point for research, not a recommendation to buy
- Not financial advice
This agent is inspired by the open-source Covenant hedge fund framework, adapted for Solana tokens and traditional tickers side by side.