PinnFeed (Pinnacle odds)

MCP server được lưu trữ cho tỷ lệ cược Pinnacle theo thời gian thực: giá trực tiếp và trước trận, tỷ lệ cược công bằng không có biên lợi nhuận và lịch sử tỷ lệ cược lúc đóng kèo dành cho Claude, Cursor và ChatGPT. Gói miễn phí (REST, độ trễ 30 giây). Không liên kết với Pinnacle.

Tài liệu

PinnFeed hosts a Model Context Protocol server so AI assistants can read live Pinnacle odds directly.

  • URL: https://pinnfeed.io/mcp
  • Transport: Streamable HTTP (spec 2025-06-18), stateless JSON responses
  • Auth: your PinnFeed API key as Authorization: Bearer <key> (or x-api-key, or ?key=<key> for clients that cannot send headers)
  • Limits: your plan's sports, data delay and rate limit. Every tool call counts as one REST request. The Free plan is delayed 30 seconds; paid plans are real time. get_closing_odds requires Builder or above.

The dashboard fills the snippets below with your key.

MCP tool calls follow your REST plan, including the Free 30-second delay. For a live + prematch Pinnacle WebSocket in your own code, try Builder WebSocket free for 3 days. Builder €99/mo (VAT incl. where applicable) after day 3 unless you cancel · card required.

Tools

ToolWhat it returns
get_accountPlan, sports, rate limit, daily quota, data delay, requests used today
list_sportsSports with live / prematch event counts and whether each is in your plan
search_eventsEvents by sport, team or league text, live or prematch
get_live_oddsIn-play events with prices (decimal, American, no-vig fair, margin, max stake)
get_prematch_oddsUpcoming events with prices, soonest first
get_eventOne event with every market
get_dropsBiggest price drops in the last 1–60 minutes
get_closing_oddsClosing odds for an event (Builder and above)

Claude Code

claude mcp add --transport http pinnfeed https://pinnfeed.io/mcp --header "Authorization: Bearer YOUR_KEY"

Claude (web and desktop)

Settings → Connectors → Add custom connector. Custom connectors cannot send headers, so put the key in the URL:

https://pinnfeed.io/mcp?key=YOUR_KEY

Cursor

~/.cursor/mcp.json (or .cursor/mcp.json in a project):

{ "mcpServers": { "pinnfeed": { "url": "https://pinnfeed.io/mcp", "headers": { "Authorization": "Bearer YOUR_KEY" } } } }

VS Code Copilot

.vscode/mcp.json:

{ "servers": { "pinnfeed": { "type": "http", "url": "https://pinnfeed.io/mcp", "headers": { "Authorization": "Bearer YOUR_KEY" } } } }

Codex CLI

export PINNFEED_KEY="YOUR_KEY"
codex mcp add pinnfeed --url https://pinnfeed.io/mcp --bearer-token-env-var PINNFEED_KEY

ChatGPT

Settings → Apps & Connectors → Advanced → enable Developer mode, then create a connector with MCP server URL https://pinnfeed.io/mcp?key=YOUR_KEY and authentication "No authentication".

Raw protocol

curl -s https://pinnfeed.io/mcp -H "Authorization: Bearer YOUR_KEY" -H "content-type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"get_drops","arguments":{"sport":"soccer","minutes":15}}}'

Tool results are JSON text. Errors (unknown sport, plan limits, rate limits) come back as a tool result with isError: true and a message the assistant can act on.

Security

For coding agents loading a skill file, see also /SKILL.md, /llms-api.txt, /AGENTS.md and the AI Catalog. Copy-paste prompts: vibe coding.

A key in a URL or a chat is a secret like any other: use a dedicated key per assistant and revoke it from the dashboard if it leaks.

REST vs WebSocket: what is actually delayed

MCP is not a WebSocket. Each tool call is one REST request and inherits that plan's delay, sports and rate limit. On Free, responses are delayed 30 seconds and REST snapshots carry X-Data-Delay-Seconds: 30 (REST docs). Paid REST is real time, but a polling client only sees a change on its next request — on average half a polling interval, plus the round trip. That is arithmetic, not a latency claim against any named vendor.

The live + prematch push transport is the Builder WebSocket (wss://pinnfeed.io/v1/stream): a snapshot when you subscribe, then delta frames as prices move (WebSocket docs). PinnFeed does not publish millisecond latency figures or a latency SLA, and does not compare latency with any named competitor. WebSocket delta frames carry a server ts in milliseconds, so you can measure your own path. How we treat latency claims: latency honesty.