MCP Server Kalshi

What mcp-server-kalshi MCP server does

The mcp-server-kalshi MCP server gives Claude Code and other MCP-compatible agent harnesses an interface to Kalshi’s prediction-market exchange. It covers the workflow from finding a market to examining its terms, reviewing market activity, checking an account, and submitting or managing orders.

Market discovery includes markets, events, and series. Research tools expose order books, candlesticks, and historical trades. For rule review, the server combines the market’s primary and secondary rules with early-close conditions, settlement sources, and series restrictions. It can also download the contract-terms PDF and extract its text for agent inspection.

Account-oriented tools provide exchange status and schedule information, balances, positions, fills, settlements, and order details. Trading actions include creating, cancelling, amending, and decreasing orders, alongside listing existing orders.

How it works

The mcp-server-kalshi MCP server translates Kalshi’s order representation into a simpler model based on an action, a YES or NO side, and a whole-cent limit price. It handles the price inversion needed when buying NO contracts versus selling YES contracts.

Public market and rule operations can run without credentials. Authenticated requests use an API key ID and an RSA private key. The request-signing scheme uses RSA-PSS with MGF1-SHA256 and sends Kalshi access-key, timestamp, and signature headers.

The default target is Kalshi’s demo environment. Set the environment explicitly to prod before using the production API. Order placement and amendment are protected by a confirmation gate: without confirm=true, the tool produces a readable preview and payload rather than sending the order.

Setup and configuration

A packaged installation can be launched with uvx mcp-server-kalshi. The README also documents Claude Desktop configuration using uvx, and a Docker option using the mcp-server-kalshi image. Local development requires Python 3.10 or newer; the documented workflow uses uv sync followed by uv run start.

Configuration is controlled with KALSHI_ENV, KALSHI_API_KEY, KALSHI_PRIVATE_KEY_PATH, and optionally BASE_URL. KALSHI_ENV accepts demo or prod and defaults to demo. BASE_URL can override the derived REST endpoint but must include /trade-api/v2. The API key and private-key path are needed only for authenticated tools.

Tools and capabilities

The documented tool groups include:

• Discovery: list or retrieve markets, events, and series.
• Research: inspect order books, candlesticks, and trades.
• Rules: consolidate market terms and extract contract PDF text.
• Exchange: check operating status and schedule.
• Portfolio: retrieve balances, positions, fills, and settlements.
• Trading: create, cancel, amend, decrease, list, and retrieve orders.

Limitations and notes

Kalshi does not provide free-text search through the API, so list_markets filters serve as the discovery mechanism. Production trading is possible when configured for prod, but the default is the sandbox environment. Credentials are not needed for public data, yet account and order functions depend on valid Kalshi authentication material. The confirmation requirement prevents create and amend operations from placing orders until the agent explicitly confirms them.