AI agent developer

You’re building agents that need prediction-market context — research assistants, signal bots, position monitors, alert systems. The Predexon MCP server gives any MCP-aware model access to tools spanning every data surface, behind one drop-in package. You’ll be ready to:

Install the MCP server in any major agent runtime
Compose Predexon tools with your own
Build the five most common agent patterns (research, monitor, alert, copy-trade, wallet analysis)

60-second install

Pick your runtime, paste the config, restart.

Claude Desktop
Claude Code
Cursor
Codex

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "predexon": {
      "command": "npx",
      "args": ["-y", "predexon-mcp"],
      "env": { "PREDEXON_API_KEY": "your-key" }
    }
  }
}

Restart Claude Desktop. Ask: “List the top 5 Polymarket markets by volume.”

claude mcp add predexon npx predexon-mcp

Edit ~/.cursor/mcp.json (global) or .cursor/mcp.json (per-project):

{
  "mcpServers": {
    "predexon": {
      "command": "npx",
      "args": ["-y", "predexon-mcp"],
      "env": { "PREDEXON_API_KEY": "your-key" }
    }
  }
}

Edit ~/.codex/config.toml:

[mcp_servers.predexon]
command = "npx"
args = ["-y", "predexon-mcp"]
[mcp_servers.predexon.env]
PREDEXON_API_KEY = "your-key"

Full config matrix (VS Code, Cline, etc.) on the MCP Server page.

The tools at a glance

Category	Count	What it covers
Polymarket markets	9	Lists, events, prices, candles, volume, OI, orderbooks
Polymarket trading	5	Trades, activity, positions, P&L
Polymarket smart-money	12	Profiles, leaderboards, top holders, smart-activity, cohort stats
Kalshi	3	Markets, trades, orderbooks
Binance	2	Crypto reference ticks + candles
Limitless / Opinion / Predict.Fun	6	Markets + orderbooks per venue

Full tool list with descriptions on the MCP Server page.

Recipe 1 — Market research assistant

Use case: “What’s happening in election markets right now?” Agent surveys, summarizes, cites. System prompt sketch:

You are a prediction-market research assistant. When asked about a topic:
1. Use `list_polymarket_markets` to find relevant markets (filter by status="open").
2. For markets with $1M+ volume, fetch `get_polymarket_candlesticks` for the last 7 days
   to identify price moves.
3. Use `get_polymarket_smart_money` on any market where price moved >5% to check if
   profitable wallets are buying or selling.
4. Summarize: which markets matter, what's the price action, what is smart money doing,
   what events are upcoming. Always cite the market slug and source endpoint.

The agent will compose tools without further prompting. Add list_kalshi_markets if you want it to cite Kalshi prices side-by-side.

Recipe 2 — Position monitor

Use case: Daily digest of how a wallet’s portfolio moved overnight. System prompt sketch:

Run daily for wallet 0x.... :
1. `get_polymarket_positions` to enumerate current positions.
2. For each position, `get_polymarket_price` to compute mark-to-market vs cost basis.
3. `get_polymarket_pnl` for the period since last run.
4. Output a digest: wins, losers, biggest mover, total P&L delta, anything that
   resolved overnight.

Pair with a cron tool (most agent runtimes have one) for fully autonomous reporting.

Recipe 3 — Smart-money alerter

Use case: Real-time alert when profitable wallets enter a market you care about. This one streams instead of polling. Agent runtimes that support background MCP tool use (Claude Code, custom builds) can call:

Every 5 minutes:
1. `get_polymarket_smart_activity` to see which markets smart money entered.
2. Filter to markets matching my watchlist (keywords: "election", "fed", "btc").
3. For each new entry, fetch `get_polymarket_smart_money` to see net positioning.
4. If net buy >$50k in the last hour, alert me with market name, smart-money flow,
   current price.

Cheaper agents can run this on a 15-min cadence with get_polymarket_smart_activity alone.

Recipe 4 — Copy-trade watcher

Use case: Track a high-performing wallet’s fills and report them the moment they land.

Watching wallet 0xSmartTrader on Polymarket:
1. `get_polymarket_wallet_profile` once at startup to verify they're still profitable.
2. Every minute: `get_polymarket_trades` filtered by user=0xSmartTrader since last check.
3. For each new trade: report market, side, size, and price, plus the wallet's
   current position in that market.
4. If I ask, compute what a mirrored position scaled to my bankroll would look like.

Execution is up to you and your own venue integration — see Copy-trade cookbook recipe for the production signal pipeline.

Recipe 5 — Wallet deep-dive analyst

Use case: Full due-diligence report on any wallet, on request.

On request for wallet 0xABC:
1. `get_polymarket_wallet_profile` — headline stats, win rate, volume.
2. `get_polymarket_wallet_pnl` — realized + unrealized P&L over time.
3. `get_polymarket_wallet_positions` — what they hold right now, by market.
4. Summarize: are they profitable, in what categories, over what period,
   and what are they betting on today?

Batch multiple wallets with get_polymarket_wallet_profiles (up to 20 per call) when ranking a leaderboard.

Tips for building good agents on Predexon

Prefer smart-money tools over raw trades

get_polymarket_smart_money and get_polymarket_smart_activity already filter to profitable wallets. Using them keeps your context window small and signal-to-noise high vs. paginating the raw trades endpoint.

Bound the wallet space before profile lookups

get_polymarket_wallet_profiles (batch) accepts up to 20 wallets per call. Have the agent batch lookups when summarizing a leaderboard.

Watch for stale data on resolved markets

Markets that resolved 2+ weeks ago may not appear in default queries. Add archived=true if your agent needs to reason about historical resolutions.

Don't combine wildcard WS subscriptions with the MCP server

The MCP server uses REST. For wildcard firehose streaming, connect to the WebSocket directly — it’s not a fit for chat-loop agent runtimes.

What you should read next

Full MCP server reference

Complete tool list with descriptions, all runtime configurations.

Cookbook

Production-grade recipes you can crib from.

Data & Signals overview

The REST endpoints behind every MCP tool.

Builder Architecture Guide

If your agent places trades — account model, custody, fee monetization.

Welcome

Guides for your use case

Cookbook

Core concepts

Account & access

Integrations

60-second install

The tools at a glance

Recipe 1 — Market research assistant

Recipe 2 — Position monitor

Recipe 3 — Smart-money alerter

Recipe 4 — Copy-trade watcher

Recipe 5 — Wallet deep-dive analyst

Tips for building good agents on Predexon

What you should read next

Full MCP server reference

Cookbook

Data & Signals overview

Builder Architecture Guide

​60-second install

​The tools at a glance

​Recipe 1 — Market research assistant

​Recipe 2 — Position monitor

​Recipe 3 — Smart-money alerter

​Recipe 4 — Copy-trade watcher

​Recipe 5 — Wallet deep-dive analyst

​Tips for building good agents on Predexon

​What you should read next

Full MCP server reference

Cookbook

Data & Signals overview

Builder Architecture Guide

60-second install

The tools at a glance

Recipe 1 — Market research assistant

Recipe 2 — Position monitor

Recipe 3 — Smart-money alerter

Recipe 4 — Copy-trade watcher

Recipe 5 — Wallet deep-dive analyst

Tips for building good agents on Predexon

What you should read next