pmxt-mcp
MCP server that provides a unified prediction market API for multiple venues like Polymarket and Kalshi, allowing AI agents to discover markets, fetch order books, and execute trades through a single interface.
README
@pmxt/mcp
MCP server that exposes the PMXT unified prediction market API as tools for Claude and other AI agents.
One tool per API method. Same interface regardless of venue -- Polymarket, Kalshi, Limitless, Probable, Baozi, Myriad, Opinion, Metaculus, Smarkets, and more.
Quick start
Add to your Claude Desktop config (claude_desktop_config.json):
{
"mcpServers": {
"pmxt": {
"command": "npx",
"args": ["-y", "@pmxt/mcp"],
"env": {
"PMXT_API_KEY": "pmxt_live_..."
}
}
}
}
Or run directly:
PMXT_API_KEY=pmxt_live_... npx @pmxt/mcp
Get an API key at pmxt.dev/dashboard.
Modes
The MCP server doesn't run prediction market logic itself -- it forwards every tool call to a PMXT API server over HTTP. Where that server lives depends on the mode:
Hosted -- Set PMXT_API_KEY and the server calls https://api.pmxt.dev. No local setup required; the hosted service manages exchange connections, caching, and rate limits for you.
Local (sidecar) -- If no API key is set, the server assumes you're running the PMXT core server locally on http://localhost:3847. This is useful for development, self-hosting, or when you want full control over the runtime. See the pmxt core repo for how to run the server.
You can point at any PMXT-compatible server by setting PMXT_API_URL explicitly.
Environment variables
| Variable | Description |
|---|---|
PMXT_API_KEY |
API key for the hosted PMXT service |
PMXT_API_URL |
Override the API base URL (defaults based on mode) |
Tools
Every tool requires an exchange parameter (e.g. polymarket, kalshi, limitless). Read-only tools are safe to call freely. Order-related tools (createOrder, submitOrder, cancelOrder) require explicit user confirmation -- they spend real money.
Market discovery: fetchMarkets, fetchMarketsPaginated, fetchEvents, fetchEvent, fetchMarket
Order book & pricing: fetchOrderBook, fetchTrades, fetchOHLCV, getExecutionPrice, getExecutionPriceDetailed
Trading: buildOrder, createOrder, submitOrder, cancelOrder
Account: fetchBalance, fetchPositions, fetchOpenOrders, fetchClosedOrders, fetchAllOrders, fetchOrder, fetchMyTrades, loadMarkets
How it works
The server translates MCP tool calls into HTTP requests to the PMXT REST API:
- Agent calls a tool (e.g.
fetchMarkets) with flat{ exchange, limit, query }input - The server reconstructs positional arguments from the flat input using embedded arg specs
- Sends
POST /api/{exchange}/{method}with{ args: [...] }to the PMXT API - Returns the JSON response to the agent
Auto-generation pipeline
The tool definitions in src/generated/tools.ts are not hand-written. They are generated from the PMXT core OpenAPI spec by scripts/generate-tools.cjs.
The full pipeline runs automatically on every PMXT release:
- A new version tag (
v*) is pushed to the pmxt core repo - The
sync-mcp.ymlGitHub Actions workflow triggers - It clones this repo, copies the latest spec files from core into
spec/:core/src/server/openapi.yaml-- full API spec (endpoints, parameters, response schemas)core/src/server/method-verbs.json-- HTTP verb and positional argument metadata per method
- Runs
node scripts/generate-tools.cjsto regeneratesrc/generated/tools.ts - Bumps
package.jsonto match the core version - Commits, tags, and pushes to this repo
- Publishes to npm with
npm publish --provenance --access public
What the generator does:
- Reads both spec files
- Skips streaming/internal methods (
watchOrderBook,close,healthCheck, etc.) - Flattens complex parameters into flat MCP tool input schemas
- Embeds
ArgSpecmetadata so the server can reconstruct positional args at runtime - Adds annotations (
readOnlyHint,destructiveHint,idempotentHint) per tool - Marks order-related tools with a
credentialsinput property
To regenerate locally:
npm run generate
Development
npm install
npm run generate # regenerate tools from spec/
npm run build # compile TypeScript
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.