Flash Props API
Live sports betting player props MCP server covering NBA, MLB, NFL, NHL, NCAA, and soccer. Unified from real sportsbooks into one REST API and a real MCP server (Streamable HTTP). Free tier, no card required.
README
Flash Props API
Sellable, agent-native sports-betting player-props data API by Flash AI Solutions. Unifies the props board into one clean REST + MCP feed — pre-game props for MLB, NFL, NBA, NHL, NCAA, soccer, tennis, and esports (CS2, Valorant, Dota 2, Call of Duty). American odds. Coverage varies by sport, season, and upstream availability.
Modeled on what Unusual Whales did: a real product with token auth, tiered rate limits, Stripe billing, an OpenAPI spec, a public storefront, and an MCP server so AI agents can consume it directly.
Stack
- Hono +
@hono/zod-openapi— routes and auto-generated OpenAPI 3.1 spec - Scalar — interactive API reference at
/docs - Drizzle + better-sqlite3 — API keys (HMAC-hashed) + usage counters
- Stripe — Checkout + webhook subscription lifecycle
- @hono/mcp + @modelcontextprotocol/sdk — Streamable-HTTP MCP server at
/mcp - Data: free upstreams only, so it costs $0 to run. (Underdog is the live source today; a Bovada adapter exists for live in-game lines but currently returns empty from the VPS.)
Endpoints
| Method | Path | Notes |
|---|---|---|
| GET | / |
Public landing page (storefront + pricing) |
| GET | /docs |
Scalar API reference |
| GET | /openapi.json |
OpenAPI 3.1 spec |
| GET | /skill.md, /llms.txt |
Agent-discovery files |
| GET | /health |
Liveness |
| GET | /api/v1/sports |
Sports + tier access |
| GET | /api/v1/games?sport= |
Today's games |
| GET | /api/v1/games/{eventId}/props?stats= |
Props for one game |
| GET | /api/v1/props?sport=&stat=&limit= |
Market-wide scan ("flow" feed) |
| GET | /api/v1/me |
Key, tier, usage |
| ALL | /mcp |
MCP server (tools: list_sports, list_games, get_game_props, scan_props, find_game) |
| GET | /billing/free |
Self-serve free key |
| GET | /billing/checkout?tier=starter|pro |
Stripe Checkout |
| GET | /billing/success |
Provision + show key after payment |
| POST | /billing/webhook |
Stripe subscription lifecycle |
Auth: Authorization: Bearer <key> (also X-API-Key: or ?api_key=).
Tiers
src/config/tiers.ts — Free (in-season sport, 250/day, 25-row scans) · Starter ($15,
all active sports, 10k/day, 100-row scans) · Pro ($39, all sports, 100k/day, 500-row
scans) · Enterprise (custom). Tiers differ by request volume, scan size, and sport
breadth — every tier serves the same pre-game snapshot (no realtime/delayed split).
Run locally
npm install
npm run db:seed # prints a PRO + FREE test key (shown once)
npm run dev # tsx watch on :3860
Then: curl -H "Authorization: Bearer <key>" "http://localhost:3860/api/v1/props?sport=mlb&limit=10"
Build & deploy (VPS)
CI/CD: push to main → GitHub Actions (.github/workflows/deploy.yml) typechecks,
builds, ships dist/ to the VPS over SSH, and pm2 restarts. No manual steps.
Secrets: VPS_SSH_KEY (dedicated ed25519 deploy key), VPS_HOST, VPS_USER.
Production runs the compiled output (node dist/index.js) — not tsx — to stay
independent of Node/tsx versions. Manual deploy (if ever needed) = build locally, ship the artifact:
npm run build # tsc -> dist/ (NodeNext, .js extensions required)
# tar excluding node_modules/.git/data/.env (anchor data exclude: --exclude=./data
# so dist/data is NOT stripped), scp to /opt/flash-props-api, extract
ssh root@5.78.189.124 'cd /opt/flash-props-api && npm install --omit=dev && pm2 restart flash-props-api'
- pm2 process:
flash-props-apion port 3862 (pm2 start dist/index.js) - Fronted by Caddy:
api.flashodds.live { reverse_proxy localhost:3862 } - Env in
/opt/flash-props-api/.env(mode 600): PORT, PUBLIC_BASE_URL, API_KEY_SECRET, DATABASE_URL, Stripe keys.
To enable billing: set STRIPE_SECRET_KEY and STRIPE_WEBHOOK_SECRET in .env and
restart. Plan prices come from src/config/tiers.ts (single source of truth), so no Stripe
price IDs are needed. Until then, checkout returns a friendly 503.
Layout
src/
index.ts entry: mounts everything, serves openapi + docs
env.ts zod-validated env
config/ tiers.ts, sports.ts
data/ underdog.ts, bovada.ts, props.ts (facade), types.ts
auth/ keys.ts (HMAC), middleware.ts
rate-limit/ limiter.ts (per-minute), middleware.ts (+ per-day quota)
db/ schema.ts, client.ts (self-bootstrapping), usage.ts, seed.ts
routes/ v1.ts (API), meta.ts (landing + skill.md), gating.ts
openapi/ schemas.ts (zod -> spec)
mcp/ server.ts (tools), http.ts (streamable transport)
billing/ stripe.ts, provision.ts, routes.ts
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。