Nephyr Risk

clm-studios/nephyr-risk

You'd reach for this when you're actively trading on prediction markets and need programmatic risk management beyond gut feel. It exposes Kelly criterion position sizing so you're not over-betting profitable edges, EV calculations to quantify expected value across positions, and drawdown protection to keep losing streaks from wiping you out. Useful if you're building trading bots or decision tools for platforms like Manifold, Polymarket, or Metaculus where you need systematic guardrails on bet sizing. The stdio transport makes it straightforward to pipe market odds and bankroll data through for position recommendations.

tools: 7
transport: STDIO
last updated: Apr 1, 2026

View on GitHub

Tools

7 tools

Public tool metadata for what this MCP can expose to an agent.

calculate_kelly_sizeCalculate the optimal position size for a prediction market trade using the Kelly criterion (quarter-Kelly by default). Returns the recommended dollar amount to risk, the raw Kelly fraction, and whether the size was capped by the bankroll-percentage limit.6 params

Calculate the optimal position size for a prediction market trade using the Kelly criterion (quarter-Kelly by default). Returns the recommended dollar amount to risk, the raw Kelly fraction, and whether the size was capped by the bankroll-percentage limit.

Parameters* required

bankrollnumber

Available capital in USD.

entry_pricenumber

Current market price for the YES contract (0 < price < 1).

kelly_fractionnumber

Scaling multiplier for full Kelly (default 0.25 = quarter-Kelly).default: 0.25

max_position_pctnumber

Hard cap as % of bankroll per trade (default 5.0%).default: 5

min_sizenumber

Minimum position size in USD (default $1.00).default: 1

probabilitynumber

Your model's probability of the YES outcome (0 < p < 1).

calculate_evCalculate the expected value of a YES contract after platform fees. Returns EV, edge (model prob − market price), fee per leg, round-trip fee, and whether EV is positive. Supports Polymarket (2% flat) and Kalshi (7% × P × (1−P)) fee models.3 params

Calculate the expected value of a YES contract after platform fees. Returns EV, edge (model prob − market price), fee per leg, round-trip fee, and whether EV is positive. Supports Polymarket (2% flat) and Kalshi (7% × P × (1−P)) fee models.

Parameters* required

market_pricenumber

Current YES price on the exchange (0 < price < 1).

model_probabilitynumber

Your model's probability of YES (0 < p < 1).

platformstring

Fee model: 'polymarket' or 'kalshi'.one of polymarket · kalshidefault: polymarket

calculate_feeCalculate the fee per leg for a given platform and contract price. Polymarket: flat 2% of price. Kalshi: 7% × P × (1−P).2 params

Calculate the fee per leg for a given platform and contract price. Polymarket: flat 2% of price. Kalshi: 7% × P × (1−P).

Parameters* required

platformstring

Fee model: 'polymarket' or 'kalshi'.one of polymarket · kalshi

pricenumber

Contract price as a probability (0 < price < 1).

check_risk_statusCheck current risk management status. Returns ACTIVE (safe to trade), PAUSED (daily loss limit hit — resume tomorrow), or SHUTDOWN (max drawdown exceeded — manual review required). Also returns warnings when approaching limits.5 params

Check current risk management status. Returns ACTIVE (safe to trade), PAUSED (daily loss limit hit — resume tomorrow), or SHUTDOWN (max drawdown exceeded — manual review required). Also returns warnings when approaching limits.

Parameters* required

bankrollnumber

Current bankroll in USD.

daily_loss_limit_pctnumber

Daily loss ceiling as % of peak bankroll (default 10%).default: 10

daily_pnlnumber

Today's realised P&L in USD (negative = loss).

drawdown_limit_pctnumber

Drawdown ceiling as % of peak bankroll (default 25%).default: 25

peak_bankrollnumber

Highest bankroll ever reached in USD.

validate_tradePre-trade risk check. Projects the worst case (total loss of proposed_size) and checks whether it would breach daily loss or drawdown limits. Returns approved/rejected with a reason and projected percentages.6 params

Pre-trade risk check. Projects the worst case (total loss of proposed_size) and checks whether it would breach daily loss or drawdown limits. Returns approved/rejected with a reason and projected percentages.

Parameters* required

bankrollnumber

Current bankroll in USD.

daily_loss_limit_pctnumber

Daily loss ceiling as % of peak bankroll (default 10%).default: 10

daily_pnlnumber

Today's realised P&L in USD (negative = loss).

drawdown_limit_pctnumber

Drawdown ceiling as % of peak bankroll (default 25%).default: 25

peak_bankrollnumber

Highest bankroll ever reached in USD.

proposed_sizenumber

Dollar amount of the proposed trade.

get_exposureSummarise portfolio exposure relative to bankroll. Returns total exposure %, largest position %, number of positions, and a concentration risk flag (triggered when any single position exceeds 10% of bankroll).2 params

Summarise portfolio exposure relative to bankroll. Returns total exposure %, largest position %, number of positions, and a concentration risk flag (triggered when any single position exceeds 10% of bankroll).

Parameters* required

bankrollnumber

Total capital in USD.

positionsarray

List of position objects. Each must have a 'size' key with the dollar amount at risk. Example: [{"size": 50.0}]

simulate_drawdownSimulate a hypothetical sequence of losses against a starting bankroll. Shows when daily loss and drawdown limits would trigger, and reports the maximum drawdown percentage reached.4 params

Simulate a hypothetical sequence of losses against a starting bankroll. Shows when daily loss and drawdown limits would trigger, and reports the maximum drawdown percentage reached.

Parameters* required

bankrollnumber

Starting bankroll in USD.

daily_loss_limit_pctnumber

Daily loss ceiling as % of peak bankroll (default 10%).default: 10

drawdown_limit_pctnumber

Drawdown ceiling as % of peak bankroll (default 25%).default: 25

loss_sequencearray

List of loss amounts in USD (positive = money lost, negative = gain). E.g. [30, 40, 50] simulates three losses.