Build with AI

Sports odds API for AI agents

An agent should resolve the event, fetch a scoped response, and preserve the evidence that supports every price. That gives the user a quote they can check.

TheRundown Product API gives an application direct access to source-backed sports data. The public OpenAPI specification describes the HTTP contract. For an agent client, the official TheRundown Data MCP repository provides a local, read-only stdio server with a deliberately small tool set. It is not a hosted data MCP service.

The remote documentation MCP at https://docs.therundown.io/mcp searches documentation only. Download the local data server from the official 0.2.1 source ZIP when an agent needs Product API data.

Keep the key in the environment

Store the Product API key in the server environment or a client secret mechanism. Send it in the X-TheRundown-Key request header. Never put a real key in a prompt, a URL, source control, or a client bundle. The local MCP server reads THERUNDOWN_API_KEY from its own process environment and does not accept credentials as tool input.

Use the smallest useful tool set

The local server exposes six read-only tools: list_sports, list_affiliates, list_markets, list_events, get_main_lines, and list_futures. Start with current catalogs, select an exact event ID from the returned event list, then request open main lines for that event. Futures and WebSocket access require Ultra or higher. Retired affiliate 27 remains excluded.

Start with a bounded request

This first conversation is intentionally narrow. It discovers the relevant catalogs, uses a UTC date, selects a returned event ID, and keeps the requested market and affiliate filters consistent.

First conversation
Use TheRundown to list current sports and affiliates. Find MLB (sport 3).
For today's UTC date, list events with market_ids [1,2,3]
and affiliate_ids [19,23]. Select an event ID from that response and
call get_main_lines with the same filters.
Show the source URL, each book's line value and price updated_at,
and the returned usage headers. Explain empty results without inventing odds.

Preserve evidence with the answer

The local data MCP result envelope retains a credential-free source_url, retrieved_at, allowed usage headers, and data. Individual prices retain updated_at. Retrieval time tells you when the MCP request completed. It does not prove that a price is fresh. The Product API can return public source_id and affiliate_source_ids fields for source mapping; the curated local MCP projection does not include them.

Different books can publish different main-line values. Keep sportsbook, prediction-market, and exchange quotes distinct. An empty, closed, stale, unauthorized, or unavailable response should stay explicit. It is not a reason to fill a gap with remembered or invented odds.

Account for plan limits before scaling

Every Product API request follows the key's entitlement, delay, coverage, and data-point billing. Read-only local MCP snapshots can consume data points. Inspect returned usage headers, keep filters bounded, and respect Retry-After on rate limits. The local MCP returns REST snapshots, not a hosted connection or an automatic retry loop.

Continue with the contract

Read the OpenAPI specification, then use the Build with AI guide for local MCP setup and the quickstart for direct API requests.