RESEARCH ENGINE FOR PREDICTION MARKETS
Trade-ready research.
On demand. For any market.
Bettors ship AI apps without research. Horizontal tools return prose. bagprint returns structured signals synthesized per request by a pipeline that forks itself for your market.
Card required. Metered from $5/run.
Why AI-built betting apps lose money.
Calibrated to prices, not the world.
AI ships the app; nobody does the research.
A real pipeline is weeks of infra.
Crawlers, source ranking, multi-agent reasoners, evals, backtests.
Horizontal deep-research returns prose.
ChatGPT / Perplexity / Parallel don't understand contract structure, liquidity, settlement.
Incumbents won't bother.
Prediction markets too small for equities/crypto — that's the gap.
What bagprint returns that they don't.
| bagprint | ChatGPT Deep Research | Perplexity | Parallel | |
|---|---|---|---|---|
| Output shape | Structured JSON | Prose | Prose | Prose |
| Tuned to prediction markets | Yes | —No | —No | —No |
| Position + confidence + size | Yes | —No | —No | —No |
| Deep-link to bet | Yes | —No | —No | —No |
| Per-request pipeline synthesis | Yes | —No | —No | —No |
How a request becomes an artifact.
Name the market
Kalshi / Polymarket / HyperLiquid id. Pass model tier + risk.
The harness plans
A research plan tuned to that market.
It forks and runs
Forks deep-research base, generates per-request code on Lambda.
You get the artifact
Position, confidence, trace, sources, deep-link.
MVP alert-only. No auto-execute. Kalshi key read-only.
See what comes back.
{"ticker": "KXNBA-LAL-25M03","market": "NBA · Lakers vs Celtics","position": "YES","entry_price": 0.62,"size_band": "2.0%","confidence": 0.72,"confidence_drivers": [ "Lakers 7-2 last 9 home", "LeBron 32.4 PPG vs Celtics this season" ],"sources": [ {"url": "https://www.espn.com/nba/...","retrieved_at": "2025-03-14T18:32:00Z" }, {"url": "https://www.rotowire.com/...","retrieved_at": "2025-03-14T18:28:00Z" } ],"deep_link": "https://kalshi.com/markets/KXNBA-LAL-25M03"}One engine. Two ways to use it.
Pick a Kalshi market, connect read-only key, set bankroll & risk. We push alerts you tap to place.
Start a trader →Metered. No subscription.
Hobby
Fixed DeepSeek V4. Target ≤ $2.
- Structured JSON artifacts
- Source citations with timestamps
- Deep-link to place on Kalshi
Premier
OpenRouter passthrough + 30%. Pick the model.
- Everything in Hobby
- Choose any model tier
- Priority queue access
Backtest add-on · priced after first 10 backtests reveal real unit cost.
“Rob wrote an NBA trading app for Kalshi using AI. Shipped fast, wins ~40%. He doesn't need a better app — he needs a better research pipeline feeding the app. That's bagprint.”