MatchMesh: AI sports data infra on Injective. Empowering AI agents with MCP tools, Agent Skills, x402 USDC micropayments & CCTP cross-chain settlement.
MatchMesh | AI Infrastructure & Pay-Per-Use Data Protocols for World Cup Agents
MatchMesh is an MCP-native, pay-per-use football data and AI orchestration platform built natively on Injective. It transforms raw sports data into an agent-ready cognitive layer, enabling autonomous AI agents and developers to query real-time World Cup statistics, lineups, live scores, and match events through standardized protocols — all settled via cryptographic USDC micropayments.
The Problem: Traditional sports APIs are designed for web browsers, not AI agents. They require expensive subscription plans, lack machine-readable schemas, and impose rigid rate limits. There is no native way for an LLM-powered agent to discover, call, and pay for premium data endpoints autonomously.
Our Solution: MatchMesh exposes 12 registry-backed Model Context Protocol (MCP) tools and 8 pre-built Agent Skills (multi-turn reasoning pipelines) that any MCP-compatible client or LLM agent can connect to instantly. Premium tools are metered on a per-call basis using the x402 cryptographic challenge-response protocol, and users fund their accounts through Circle CCTP cross-chain USDC settlement on Injective.
Injective Technologies Integrated:
x402 Micropayments : Premium API endpoints return HTTP 402 challenges. Clients sign challenges using HMAC-SHA256 keys, and the server verifies signatures, validates receipts, and debits the user's ledger balance automatically.
Circle CCTP Settlement : Users deposit USDC into their platform ledger through cross-chain burn-and-mint transfers. The backend includes double-deposit protection and replay-attack checks.
MCP Server : A Fastify-powered JSON-RPC server hosts 12 standardized MCP tools (live scores, player stats, team standings, match events, lineups, referees, and search). Any MCP client connects instantly without custom parsing.
Agent Skills : 8 high-level reasoning skills (Match Summary, Fantasy Assistant, Player Comparison, Team Comparison, Tactical Analysis, VAR Explainer, Commentary Generator, Stats Explainer) compose raw MCP tools into multi-step workflows with Zod-validated structured outputs.
Key Features:
• Keplr Wallet authentication on Injective (injective-1 mainnet / injective-888 testnet) with EVM hex address resolution
• Published TypeScript SDK (matchmesh-sdk) on NPM with built-in x402 challenge auto-signing
• Vercel AI SDK and LangChain framework adaptors for instant agent integration
• Glassmorphic Next.js 16 developer platform with interactive playground, tool catalog, billing history, and real-time analytics
• Hexagonal architecture with provider-agnostic data adapters enforced by CI
• PostgreSQL system of record, Redis cache layer (5s live TTL), and Docker Compose containerization
How Users Interact:
Tech Stack:
Frontend: Next.js 16, React 19, Tailwind CSS v4
Backend: Fastify, TypeScript, TSyringe
Database: PostgreSQL 16, Redis 7, Prisma ORM
Infrastructure: Docker Compose, Turborepo
SDK: matchmesh-sdk
<p>I built the entire <strong>MatchMesh</strong> ecosystem from scratch during this hackathon. Here is the breakdown of what was designed, developed, and deployed:</p><ul><li><p><strong>Injective Native Architecture:</strong> Integrated Keplr Wallet to resolve EVM-compatible addresses on Injective for seamless authorization.</p></li><li><p><strong>x402 Micropayment Protocol:</strong> Designed and implemented a custom HTTP 402 (Payment Required) middleware that intercepts premium data calls, issues an HMAC-SHA256 signature challenge, and debits the user's ledger for sub-cent usage.</p></li><li><p><strong>Circle CCTP Integration:</strong> Developed a cross-chain USDC funding handler, allowing developers to deposit mock USDC and settle it into their MatchMesh ledger instantly.</p></li><li><p><strong>Ingestion Worker & Database:</strong> Designed a PostgreSQL schema using Prisma ORM. Developed a background worker to ingest live football match events and league data.</p></li><li><p><strong>Model Context Protocol (MCP) Tools:</strong> Created standardized MCP tools and Agent Skills (like VAR Explainer, Match Schedules, Player Statistics) validated via strict Zod schemas.</p></li><li><p><strong>Published SDK (</strong><code data-inline="true" spellcheck="false">matchmesh-sdk</code><strong>):</strong> Developed and published the official <code data-inline="true" spellcheck="false">matchmesh-sdk</code> package on NPM. Added out-of-the-box adapters for LangChain and Vercel AI SDK.</p></li><li><p><strong>Developer Portal Dashboard:</strong> Built a sleek Next.js dark-mode dashboard showcasing real-time API logs, billing ledger transactions, CCTP deposits, and an interactive tool playground with a UUID discovery flow.</p></li></ul>
<p><span style="color: rgb(204, 204, 204)"><strong>MatchMesh is currently a hackathon project built for the Injective Global Cup Hackathon. No external funding has been raised. We are open to grants, incubation programs, and strategic partnerships to scale the platform post-hackathon.</strong></span></p>