PropLine for AI Agents

The sports odds API your AI agent
already knows how to use.

One npx line gives Claude, ChatGPT, or any MCP client live player props and game lines across 18 books and 54+ sports — plus the thing agents actually need to be trustworthy: every prop graded against the real box score, so answers can be verified, not guessed.

Install the MCP server Get a free API key

Free tier: 1,000 requests/day, no card. Works with Claude Desktop, Claude Code, ChatGPT, Cursor, and any MCP client.

30-second setup

The propline-mcp server (listed in the official MCP Registry) exposes 21 tools covering the full REST surface — odds, props, resolution, +EV, best line, trends, futures, movement, exports.

Claude Desktop / any MCP client

{
  "mcpServers": {
    "propline": {
      "command": "npx",
      "args": ["-y", "propline-mcp"],
      "env": { "PROPLINE_API_KEY": "YOUR_API_KEY" }
    }
  }
}

Claude Code

claude mcp add propline -e PROPLINE_API_KEY=YOUR_API_KEY -- npx -y propline-mcp

No MCP? Point any model at prop-line.com/llms-full.txt and it can write REST calls directly — the whole API reference fits in one context window.

What your agent can answer

Real prompts, and the tools the agent reaches for. Every answer is backed by live data or graded outcomes — not the model's training-data memory of last season.

“What's tonight's MLB +EV slate?”

propline_list_events → propline_get_event_ev per game. Pinnacle-anchored no-vig fair lines, EV% per book at the same line.

“Did Aaron Judge's home-run prop hit last night?”

propline_get_event_results — the actual stat from the official box score plus the won/lost/push verdict. The agent verifies instead of guessing.

“Which book pays the most on this line?”

propline_get_best_line — best American price per (market, player, line) across every comparable book, all_prices ranked best-first.

“Is this player's Over trend real?”

propline_get_player_trends — L5/L10/L20/L50 over/under splits + current streak per market, derived from graded outcomes, not scraped narratives.

“How has this line moved since open?”

propline_get_event_movement — opening vs latest per (book, market, outcome) plus a steam[] array flagging synchronized cross-book moves.

“Backtest my model against closing lines.”

propline_get_odds_closing per event, or the bulk resolved-props CSV export with closing_price per row.

Built agent-first

Most odds APIs assume a human is reading the error message. PropLine assumes a program — often a model — is.

A format your model already knows

Responses are the-odds-api compatible (bookmakers[] → markets[] → outcomes[]) — the odds-API shape LLMs have seen most in training. Agents write working PropLine code zero-shot; existing the-odds-api integrations swap with one base-URL change.

Ground truth an agent can verify against

Every prop is graded against the official box score — prop resolution no other odds API offers. An agent that can check “did it hit?” gives answers, not vibes. Hallucination-resistance is a data-source feature, and resolution is ours.

llms.txt + llms-full.txt

/llms.txt is the index; /llms-full.txt is the complete API reference in one file. Paste one URL into any model's context and it has everything needed to write working code — endpoints, market keys, error semantics, common patterns.

Machine-actionable errors

Gates and throttles return a structured detail object with a stable error code, the required_tier, and the exact upgrade_url. Rate limits carry Retry-After and X-Daily-* quota headers. Your agent backs off correctly and tells you exactly what unlocks the feature — no prose parsing.

Redacted teasers, not dead ends

Free-tier calls to paid analytics (/ev, /best-line, /odds/history, /trends) return the full response structure with values nulled and an upgrade_url — a 200, not a 403. An agent can show its user exactly what upgrading unlocks before a dollar is spent.

Push, not just poll

HMAC-signed webhooks for line movement, prop resolution, and cross-book steam alerts. An agent (or the pipeline feeding it) reacts to a graded prop or a synchronized sharp move the moment it happens instead of burning requests polling.

Every gate tells the agent exactly what unlocks it

{
  "detail": {
    "error": "upgrade_required",
    "message": "Prop resolution is a paid feature — start at $9/mo Hobby. Upgrade at https://prop-line.com/pricing",
    "required_tier": "hobby",
    "upgrade_url": "https://prop-line.com/pricing",
    "docs_url": "https://prop-line.com/docs"
  }
}

The sports odds API your AI agentalready knows how to use.