Prediction Market Agent Tools
Give your AI agent access to prediction market intelligence. One line to connect. Six tools available.
MCP Server
Works with Claude Code, Cursor, Cline, Roo Code — any MCP-compatible client.
# Claude Code
$ claude mcp add simplefunctions --url https://simplefunctions.dev/api/mcp/mcp
# Cursor / Cline / Roo Code — add to MCP config:
{
"mcpServers": {
"simplefunctions": {
"url": "https://simplefunctions.dev/api/mcp/mcp"
}
}
}That's it. Your agent now has six tools:
get_contextThesis snapshot: causal tree, edges, orderbook, confidence
list_thesesAll your theses with status and confidence
inject_signalPush news, notes, or external events into a thesis
trigger_evaluationRun deep analysis with heavy model
scan_marketsSearch Kalshi markets — no API key needed
create_thesisBuild a new thesis with causal model
What your agent can do
> check recession markets for mispriced contracts
Tool: get_context f582bf76
Recession 2026 YES @ 35¢ — your thesis implies 72¢.
Edge: +37¢, spread 1¢, liquidity high.
You hold 3022 contracts at 32¢ (P&L: +$91).
Recommendation: Hold. Edge is still large. Consider
adding if price drops below 30¢.> scan oil markets for opportunities
Tool: scan_markets query="oil"
Found 12 oil-related series on Kalshi:
KXWTIMAX WTI Crude Oil Max Price 2026 vol 8.2M
KXWTIMIN WTI Crude Oil Min Price 2026 vol 3.1M
KXGAS Gas Price Above Threshold vol 1.4M
...
KXWTIMAX-26DEC31-T150 has the largest edge from your
thesis (thesis 75¢, market 38¢, edge +37).Interactive CLI Agent
For human-in-the-loop oversight. Same tools, natural language interface.
$ sf agent f582bf76
> compare my positions against current edges
⚡ get_context ✓ (0.3s)
⚡ get_positions ✓ (0.5s)
Position Avg Now P&L Edge Action
KXWTIMAX-T135 57¢ 60¢ +$23 +20 HOLD
KXRECSSNBER-26 32¢ 35¢ +$91 +37 HOLD
Unpositioned opportunities:
WTI $150 YES — 38¢ — +37 CONSIDER
Gas $4.50 Mar — 14¢ — +41 CONSIDER$ npm install -g @spfunctions/cli
$ sf agentSessions persist locally. Use /edges /tree /pos /eval for quick views.
REST API
If MCP doesn't fit your stack, use the REST API directly.
GET /api/thesis/:id/context — thesis snapshot
POST /api/thesis/:id/signal — inject signal
POST /api/thesis/:id/evaluate — trigger evaluation
POST /api/thesis/create — create thesis
GET /api/thesis — list all thesesFull docs at simplefunctions.dev/docs
How it works under the hood
You feed the system a thesis
US-Iran war persists, oil stays elevated, recession by Q3
It builds a causal tree
Structured assumptions with probabilities. Each node is verifiable.
Every 15 minutes, the heartbeat engine runs
Scans news (Tavily), refreshes prices (Kalshi + Polymarket), enriches with orderbook depth, evaluates signals, updates confidence.
Your agent reads the latest state
Via get_context — one call returns everything: causal tree, edges, orderbook, confidence, last evaluation.
Your agent injects observations
Via inject_signal — news, notes, or external events. The backend evaluates them automatically.
The agent doesn't need to search for news or fetch prices. The backend does it. The agent only needs to read, write notes, and occasionally trigger deep evaluation.
Free during beta · charge by token after 15M