MCP Server Documentation

Overview

The predictions knowledge base for AI agents. Real-time prediction-market intelligence from Kalshi and Polymarket, exposed as a Model Context Protocol server. Search markets, compare prices, surface cross-venue arbitrage opportunities, track whale signals, and pull a single-call prediction_brief on any topic.

Designed for agent reasoning, not internal ops: 13 curated tools, stable JSON contracts, a synthesis tool that collapses what used to take 5–7 separate API calls into one round-trip. Connect from Claude Code, Cursor, Windsurf, ChatGPT Apps, or any MCP-compatible client.

Connects via Streamable HTTP (MCP protocol revision 2025-03-26) and authenticates with an X-API-Key header. Free tier 100 calls/day; Pro / Intelligence tiers unlock prediction_brief, signal tools, and webhook subscriptions.

Connect

Free tier (100 calls/day) is available to every signed-in user. Pro and Intelligence add higher limits and signal tools.

Authentication

All connections require an X-API-Key header with a valid API key. Free tier 100 calls/day; Pro 10,000/day; Intelligence unlimited.

401 — Missing or invalid API key
403 — Tier required (tool requires Pro or Intelligence)
429 — Rate limit exceeded (daily or per-minute)

Claude Desktop

Open Settings → Developer → Edit Config and add to claude_desktop_config.json:

claude_desktop_config.json

{
  "mcpServers": {
    "evens-predictions": {
      "type": "http",
      "url": "https://evens.trade/mcp",
      "headers": {
        "X-API-Key": "YOUR_API_KEY"
      }
    }
  }
}

Claude Desktop speaks Streamable HTTP natively — no bridge required.

Claude Code

Run this command in your terminal to add the Evens MCP server:

Terminal

claude mcp add evens-predictions \
  --transport http \
  --header "X-API-Key: YOUR_API_KEY" \
  https://evens.trade/mcp

Claude Code natively supports the Streamable HTTP transport. The server will be available in all your conversations.

Cursor

Open Settings → MCP and add a new server, or add to your project's .cursor/mcp.json:

.cursor/mcp.json

{
  "mcpServers": {
    "evens-predictions": {
      "type": "http",
      "url": "https://evens.trade/mcp",
      "headers": {
        "X-API-Key": "YOUR_API_KEY"
      }
    }
  }
}

Other MCP clients

Any MCP client that supports the Streamable HTTP transport (MCP protocol revision 2025-03-26) can connect. POST JSON-RPC requests to the endpoint and include your API key:

HTTP endpoint + auth

URL:    https://evens.trade/mcp
Header: X-API-Key: YOUR_API_KEY

For stdio-only clients, bridge with npx mcp-remote https://evens.trade/mcp --header "X-API-Key: YOUR_API_KEY".

Tools

All tools are read-only — they mirror the REST API capabilities. Search and browse prediction market data aggregated from Kalshi and Polymarket.

Tool	Description	Key Args
search_events	Search canonical events with filters	query, sport, competition, status, start_after, start_before, limit
show_event	Get full event details with all its markets	id
search_markets	Search canonical markets with filters	query, market_type, status, event_id, sport, provider, limit
show_market	Get full market details with provider references and current prices	id
get_market_prices	Get price history for a market with time-series snapshots	id, provider, interval, start_time, end_time, limit
list_sports	List all available sports	—
list_competitions	List competitions, filterable by sport	sport, type
list_providers	List prediction market providers with counts	—
prediction_brief	Single-call topic synthesis. Returns matching markets with consensus prices, recent moves, current arbitrage, whale signals, related events, and source freshness — everything to reason about a topic in one round-trip.	topic, event_id, markets_limit
find_arbitrage	Open cross-provider arbitrage opportunities with worked-example payoffs	min_roi, limit
recent_whale_trades	Recent large positions across providers (filterable by market / event / platform)	min_size_usd, period, market_id, event_id, platform, limit
price_history	Time-series price snapshots per provider, granularity auto-selected by timeframe	market_id, timeframe, max_points_per_provider

Examples

Once connected, just ask your AI agent naturally. Here are some example prompts:

Topic-level intelligence (the killer call)

One prediction_brief call returns consensus prices, recent moves, arbitrage opportunities, whale activity, related events, and source freshness. Designed so an LLM can reason in one prompt without follow-up tool calls.

Agent prompt

"What's the consensus on the next US presidential election?"
"Brief me on the Lakers @ Warriors game tonight."
"Pull a prediction_brief for event 13324."

Find alpha — arbitrage and whale signals

Agent prompt

"Are there any prediction-market arbitrage opportunities right now with ROI over 1%?"
"Show recent whale trades over $50k on Polymarket in the last 24 hours."
"Any large positions on AI safety bill markets this week?"

Explore prediction markets

Agent prompt

# Search for events
"Find upcoming NBA events"
"Show me soccer events starting this week"

# Browse markets
"Search for markets about bitcoin"
"What player prop markets are available for basketball?"

# Get details
"Show me event 42 with all its markets"
"Show market 128 with current prices from each provider"

Compare prices across providers

Agent prompt

# Check provider coverage
"Which providers are available and how many markets do they have?"

# Compare prices
"Show me market 128 — which provider has the best yes price?"

# Historical prices
"Get hourly price history for market 128 over the last 24 hours"
"Compare daily prices from Kalshi vs Polymarket for market 128"

Browse sports and competitions

Agent prompt

# Discover what's available
"List all sports"
"Show me competitions for basketball"

# Filter by sport
"Find all open futures markets for soccer"
"What events are coming up in the NBA?"