Simple Functions

Perception. Analysis.
Execution.

SimpleFunctions is an intelligence engine for prediction markets. It watches the world, reasons about what matters, and acts when the edge is right.

9,706 markets44 MCP tools43 CLI commands15 min refresh

Perception

What is happening in the world right now?

Real-time data from prediction markets, traditional markets, and social signals — distilled into context your agent can reason over.

World State

9,706 prediction markets distilled into ~800 tokens. Calibrated by real money. One curl, no auth.

$ curl simplefunctions.dev/api/agent/world

?focus=energy,geo for deeper coverage. /delta for ~30 tokens of changes only.

Query

LLM-enhanced search across Kalshi, Polymarket, X/Twitter, and traditional markets. Ask in plain English.

$ sf query "iran oil impact"

Returns contracts, prices, sentiment, and a synthesized answer. No auth.

Monitor the Situation

Scrape any URL, analyze with any LLM, cross-reference with prediction markets, push to webhook.

$ POST /api/monitor-the-situation

Firecrawl + OpenRouter + 9,706 markets. Enrich endpoint is free.

World State
$ curl simplefunctions.dev/api/agent/world

Prediction Index  Uncertainty: 67 · Geo Risk: 85

Geopolitics
  Iran invasion          53¢ +5¢
  Hormuz <3 ships/day    95¢

Economy
  Recession 2026         35¢ +3¢
  Fed cuts before Jul    42¢ -2¢

Traditional
  WTI $127.40 +3.2%  SPY 512.40
  VIX 18.3   Gold $2,890

Analysis

Turn intuition into structure.

A thesis in plain English becomes a causal tree of testable claims, each mapped to real contracts. The system finds where the market disagrees with you — and why.

Edge Discovery
$ sf edges

WTI T150 YES    +36¢  consensus_gap
  Market 38¢ · Thesis 74¢ · Depth $2,400

Gas $4.50 YES   +41¢  attention_gap
  Market 14¢ · Thesis 55¢ · Depth $800

Recession YES   +20¢  timing_gap
  Market 35¢ · Thesis 55¢ · Depth $1,200

Track record: 67% of edges moved in thesis direction

Causal Tree

Natural language → directed graph of testable claims. Each node has a probability and maps to real contracts.

$ sf create "Oil above $100 through 2026"

Weekly augmentation adds nodes from evaluation insights. Append-only.

Edge Discovery

Where your model disagrees with the market. Four types: consensus gap, attention gap, timing gap, risk premium.

$ sf edges --json

Each edge sized against orderbook depth, scored for executable edge after spread.

What-If

Override node probabilities, see how edges change. Zero LLM cost — pure propagation.

$ sf whatif <id> --set "n1=0.1"

Test scenarios before committing capital.

Track Record

Hit rate: % of edges that moved toward thesis. Feeds back into evaluation prompt so the model calibrates.

$ sf performance

Settlement sync tracks realized P&L from settled contracts.

Execution

From edge to fill.

Declare what you want. Set conditions. The runtime handles timing. The heartbeat handles lifecycle — requotes, cancellations, fill tracking.

Intents

Declarative execution: what to trade, how much, under what conditions. Price, confidence, and time triggers.

$ sf intent buy TICKER 200 --price 40 --trigger below:38

REST: POST /api/intents. MCP: create_intent tool.

Runtime

Daemon that evaluates triggers every 30 seconds and places orders when conditions are met.

$ sf runtime start

Records fills, updates positions. sf runtime status to check.

Orderbook

Full bid/ask ladder, liquidity scoring, slippage estimation. Kalshi and Polymarket.

$ sf book KXWTIMAX-T150

GET /api/public/market/TICKER?depth=true for 5-level depth.

Intent → Runtime → Fill
$ sf intent buy KXWTIMAX-T150 200 --price 40 --trigger below:38
✓ Intent armed — waiting for trigger

$ sf runtime start
Runtime active

  11:14  WTI T150: 39¢ — waiting
  11:15  WTI T150: 37¢ — trigger!
  11:15  ✓ Filled 200 YES @ 37¢

  Cost: $74.00
  Edge at entry: +37¢

Monitoring

Runs while you sleep.

The heartbeat engine runs every 15 minutes per thesis. News scanning, price tracking, kill condition checks, adversarial search, LLM evaluation — all autonomous.

Heartbeat

Every 15 min: 3 news queries (1 adversarial), price rescan, milestone detection, kill check, LLM eval.

$ sf context <id> --json

Fully autonomous. Your agent just reads the results.

Feed

Every evaluation timestamped with confidence change and summary. Your thesis's full history.

$ sf feed --hours 24

REST: GET /api/feed?hours=24&limit=200

Delta

~50 bytes when nothing changed, full diff when something did. 100x cheaper than polling context.

$ sf delta <id>

Pattern: poll /changes, fetch /context only on change.

Agent

Built for agents, usable by humans.

Every feature is accessible via MCP (44 tools), REST API, and CLI. The interactive agent gives you natural language on top.

Interactive Agent

Natural language session with 50 tools. Slash commands for common actions.

$ sf agent [id]

/edges, /tree, /buy TICKER QTY, /eval, /switch

Skills

Reusable prompts: /morning briefing, /discipline review, /precheck argues against your trade.

$ sf agent → /morning

Create, fork, and publish skills to the community.

Telegram

Same tools, in Telegram. Auto-pushes confidence alerts. Runs locally, credentials stay on your machine.

$ sf telegram --token TOKEN

/context, /positions, /edges, /balance, /eval

Start with one curl.

Every feature works via CLI, MCP, and REST. Public endpoints need no auth.

$ curl simplefunctions.dev/api/public/context
Get API KeyFull docs →
APIMCP ToolsCLIWorld StateTakerMaker