Why SimpleFunctions

Your prediction market agent is only as good as the context it has. SimpleFunctions solves the two hardest problems in agent-driven trading: cold start and information decay.

The cold start problem

When your agent encounters a new market — say, "Will the US-Iran war end by June?" — it knows nothing. It doesn't understand:

• The causal structure — how oil prices, military deployments, and diplomatic signals relate to conflict duration
• What markets exist across Kalshi and Polymarket that touch this thesis
• Historical patterns — how similar geopolitical markets have resolved
• Key actors, timelines, and decision points that matter
• What "edge" looks like — when a market is mispriced vs. correctly priced

Without this context, your agent is guessing. It might make a technically correct API call to place a trade, but it has no basis for evaluating whether the trade is good.

Prebuilt context packages

SimpleFunctions provides prebuilt context packages for major trading domains:

Your agent loads a context package and immediately understands the domain — the key players, the causal relationships, the relevant markets, and what signals matter. No cold start. No hallucinated analysis. Real domain expertise from the first API call.

$ sf context load geopolitics/iran-us
✓ loaded: 47 causal nodes, 12 active markets, 156 signals

$ sf thesis set "US-Iran war will not end quickly"
✓ thesis anchored to geopolitics/iran-us context
✓ 8 markets matched, 3 edges detected

The information decay problem

The traditional agent workflow for prediction markets looks like this:

1. User asks agent to research a market
2. Agent searches the web
3. Agent reads 10-20 articles
4. Agent forms a view
5. Agent places a trade
6. ... time passes ...
7. New information emerges
8. Agent doesn't know about it
9. Market moves against you
10. You realize too late

The problem: step 6-10. After the initial research, your agent forgets everything. Context decays. New signals emerge — a carrier group moves, oil spikes, a diplomatic back-channel opens — and your agent has no idea because it already "finished" its research.

Thesis-anchored continuous context

SimpleFunctions flips this model. Instead of one-shot research, your thesis becomes a persistent anchor. Context accumulates around it continuously:

Thesis: "US-Iran war will not end quickly"
          
t+0h    ✓ Thesis stored, 8 markets matched
t+15m   + Signal: Oil futures +4% (auto-scanned)
t+30m   + Signal: Carrier group repositioned (news)
t+1h    + Signal: IRGC deploys coastal missiles (news)
t+1h    ⚡ Edge widened: "War ends by June" now 22¢
t+2h    + Signal: EU sanctions package drafted (news)
t+2h    + Signal: Treasury yields +18bps (market data)
t+4h    + User signal: "Back-channel talks collapsed"
t+4h    ⚡ Re-evaluation: confidence increased to 89%

Every 15 minutes, the runtime evaluates your thesis against accumulated context. Your agent never has to re-search. It never misses a signal. The context just grows — and when something matters, the agent gets pushed an update.

The combined effect

Prebuilt context + continuous updates = an agent that trades like a domain expert who never sleeps.

Free plan · charge by token after 15M