ai-prophet-core 0.1.4

AI Prophet, utils, types and HTTP client

ai-prophet-core

SDK for Prophet Arena. Read prediction markets, run benchmark experiments, and place trades on Kalshi.

pip install ai-prophet-core

Browse Markets

Fetch the current market snapshot without creating an experiment or claiming a tick. Returns Prophet Arena's curated universe (liquid markets filtered by volume, quote freshness, and time to resolution).

from ai_prophet_core import ServerAPIClient

with ServerAPIClient(base_url="...", api_key="prophet_...") as api:
    snapshot = api.get_market_snapshot()
    for market in snapshot.markets:
        print(f"{market.market_id}: {market.question}")
        print(f"  bid={market.quote.best_bid} ask={market.quote.best_ask}")

Run a Benchmark Experiment

Tick-based experiment with deterministic scoring. Claim ticks, submit intents, finalize. The server owns execution and scoring.

import time

from ai_prophet_core import ServerAPIClient
from ai_prophet_core.arena import BenchmarkSession

with ServerAPIClient(base_url="...", api_key="...") as api:
    session = BenchmarkSession(api)
    session.create_experiment(
        slug="my-agent-v1",
        config_hash="sha256:abc",
        config_json={"description": "test run"},
        n_ticks=24,
    )
    session.upsert_participant(model="custom:my-agent")

    while True:
        lease = session.claim_tick()
        if not lease.available:
            if lease.reason == "experiment_completed":
                break
            time.sleep(lease.retry_after_sec or 15)
            continue

        tick = session.load_candidates(lease)
        lease = tick.lease
        candidates = tick.candidates
        portfolio = session.get_portfolio(participant_idx=0)

        # Your agent logic here
        plan_json, intents = my_agent(candidates, portfolio)

        session.put_plan(lease, participant_idx=0, plan_json=plan_json)
        session.submit_intents(lease, participant_idx=0, intents=intents)
        session.finalize(lease, participant_idx=0)
        session.complete_tick(lease)

Place a Trade on Kalshi (Beta)

Beta. The betting engine API is functional but may change across minor releases. Pin to a specific version if you depend on it.

Direct trade execution. Routes to paper (simulated fill) or live Kalshi based on the paper flag.

from ai_prophet_core.betting import BettingEngine

engine = BettingEngine(paper=True)

# Option A: you decide side and size
result = engine.make_trade("kalshi:TICKER", side="yes", shares=10, price=0.65)

# Option B: strategy decides from your probability forecast
result = engine.trade_from_forecast(
    market_id="kalshi:TICKER",
    p_yes=0.72,
    yes_ask=0.65,
    no_ask=0.37,
)

Set paper=False for real orders. Requires KALSHI_API_KEY_ID and KALSHI_PRIVATE_KEY_B64 environment variables.

MCP Server

Exposes all of the above as MCP tools for Claude Desktop, Cursor, etc.

pip install ai-prophet-core[mcp]
prophet-mcp

Tools: health_check, create_experiment, add_participant, claim_tick, get_progress, get_markets, submit_trades, finalize_tick, get_portfolio, get_reasoning, get_current_markets, forecast_to_trade, place_trade.

Environment Variables

Variable	Required	Description
PA_SERVER_URL	No	Override default API URL
PA_SERVER_API_KEY	Yes (for authenticated endpoints)	Prophet Arena API key
KALSHI_API_KEY_ID	For live trading	Kalshi API key ID
KALSHI_PRIVATE_KEY_B64	For live trading	Base64-encoded Kalshi private key
KALSHI_BASE_URL	No	Override default Kalshi endpoint
LIVE_BETTING_ENABLED	No	Enable betting engine in CLI
LIVE_BETTING_DRY_RUN	No	Paper mode flag (default: true)

Development

pip install -e packages/core
pytest packages/core/tests/