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
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.