Gameball MCP Server

Gameball MCP Server

Unified MCP server for Gameball's documentation and API integration, enabling documentation search, code generation, and API operations for customers, transactions, orders, coupons, and more.

Category
Visit Server

README

Gameball Integrations

Unified Model Context Protocol (MCP) server providing comprehensive access to Gameball's developer documentation and API integration.

Features

This server combines two subsystems and a status tool:

Documentation Tools (7 tools)

  • No API keys required - Automatically fetches and indexes documentation from GitHub
  • Search and browse Gameball developer docs
  • Get API endpoint specifications
  • Generate code examples in multiple languages (JavaScript, Python, C#, Go, PHP, Java)
  • Access tutorials and guides

Integration Tools (71 tools)

  • Requires API keys - Make real API calls to Gameball's services
  • Customer management (create, update, query, balance, tags, notifications)
  • Transaction handling (cashback, redemption, holds, refunds, OTP)
  • Order tracking and queries
  • Coupon validation and burning
  • Batch operations
  • Configuration queries
  • Referral system
  • Leaderboards
  • Events and payments

Status Tool (1 tool)

  • gameball-status - Shows server health, subsystem status, configuration, and cache info

Total: 79 tools in one server

Installation

Prerequisites

  • Node.js >= 20
  • Clone and build the project:
git clone <repository-url> gameball-integrations
cd gameball-integrations
npm install
npm run build

Add to Your Project (recommended)

Create a .mcp.json file at the root of your project. Claude Code automatically detects this file and loads the MCP server for that workspace.

Docs only (no API keys needed):

{
  "mcpServers": {
    "gameball": {
      "command": "node",
      "args": ["/path/to/gameball-integrations/dist/index.js"]
    }
  }
}

Docs + Integration (requires API keys):

{
  "mcpServers": {
    "gameball": {
      "command": "node",
      "args": ["/path/to/gameball-integrations/dist/index.js"],
      "env": {
        "GAMEBALL_API_KEY": "your-api-key",
        "GAMEBALL_SECRET_KEY": "your-secret-key"
      }
    }
  }
}

Tip: If the MCP repo is a sibling directory, use a relative path: "args": ["../gameball-integrations/dist/index.js"]

Add Globally (all projects)

To make the tools available across all projects, add the config to ~/.claude/settings.json under the same mcpServers structure shown above.

Via Claude Code CLI

# Docs only
claude mcp add gameball node /path/to/gameball-integrations/dist/index.js

# With API keys
claude mcp add gameball node /path/to/gameball-integrations/dist/index.js \
  -e GAMEBALL_API_KEY=your-api-key \
  -e GAMEBALL_SECRET_KEY=your-secret-key

Environment Variables

Variable Required Description Default
GAMEBALL_API_KEY For integration tools Your Gameball API key ""
GAMEBALL_SECRET_KEY For sensitive operations Your Gameball secret key undefined
GAMEBALL_BASE_URL No API base URL https://api.gameball.co
GITHUB_TOKEN No GitHub token for higher rate limits undefined

Tool Categories

Documentation Tools (docs-* prefix)

All documentation tools work without API keys:

Tool Description
docs-lookup-topic START HERE - Comprehensive lookup returning tutorials, docs, and API endpoints for a topic
docs-get-doc Get full content of a specific documentation page
docs-get-api-endpoint Get detailed API endpoint specification with parameters and schemas
docs-generate-code-example Generate code examples in JavaScript, Python, C#, Go, PHP, or Java
docs-search-docs Full-text search across all documentation
docs-list-sections Browse documentation structure and available topics
docs-get-tutorial Get step-by-step tutorial content

Integration Tools (by domain)

All integration tools require GAMEBALL_API_KEY. Tools marked with šŸ”’ also require GAMEBALL_SECRET_KEY.

Authentication (1 tool)

  • generate-session-token - Generate JWT session token

Customers (21 tools)

  • create-customer, get-customer, update-customer, delete-customer
  • get-customer-balance, adjust-customer-balance šŸ”’
  • get-customer-referrals, send-customer-notification
  • get-customer-tags, add-customer-tag, remove-customer-tag
  • And more...

Transactions (13 tools)

  • cashback-transaction, redeem-points, redeem-points-otp
  • hold-points, cancel-hold, refund-transaction
  • query-transactions, reverse-transaction
  • And more...

Orders (4 tools)

  • track-order, refund-order, query-orders, cancel-order

Coupons (7 tools)

  • validate-coupon, burn-coupon, lock-coupon, unlock-coupon
  • revoke-burn-coupon, revoke-validation, list-customer-coupons

Batch Operations (9 tools)

  • batch-adjust-balance, batch-send-events, batch-cashback
  • batch-redeem, batch-refund, batch-hold, batch-cancel-hold
  • And more...

Configuration (10 tools)

  • get-action-config, get-challenge-config, get-level-config
  • get-tier-config, query-tiers, query-levels
  • And more...

Other Domains

  • Custom (2 tools): Utilities like customer count, cart tracking
  • Events (1 tool): Send customer events
  • Leaderboard (1 tool): Get leaderboard rankings
  • Payments (1 tool): Record payments
  • Referrals (1 tool): Validate referral codes

Status Tool

Tool Description
gameball-status Shows subsystem health, API key configuration, cache location, and sync status

How It Works

Startup

  1. Fetches the latest commit SHA from GitHub (single API call)
  2. Loads both subsystems in parallel:
    • Docs subsystem: Fetches .mdx files and docs.json, builds MiniSearch index
    • Integration subsystem: Fetches openapi.json independently (gzip-cached)
  3. Parses the OpenAPI spec into endpoint definitions (shared by both subsystems)
  4. Registers all tools and connects via stdio

Progress is reported to stderr during loading:

Checking for updates...
Loading...
Ready — 79 tools

Documentation Subsystem

  • GitHub Fetching: Fetches Gameball docs from GitHub with batched requests
  • Commit-SHA Cache: Validates cache with 1 API call, re-fetches only when docs change
  • Full-Text Indexing: MiniSearch with fuzzy matching and heading-based chunking
  • Stale-Cache Fallback: Uses cached docs if GitHub is unreachable

Integration Subsystem

  • Independent OpenAPI Fetch: Fetches only openapi.json (not the full docs repo)
  • Gzip Cache: Stores the OpenAPI spec compressed with atomic writes
  • Auto-Generated Tools: All 71 tools generated from OpenAPI spec at startup
  • Smart Adaptive Verbosity: Intelligent descriptions save 78% tokens (~7,000 tokens)
  • HTTP Client: Wraps Gameball API v4.0 with automatic auth header injection

Graceful Failure

All GitHub fetches have a 10-second timeout. If GitHub is unavailable:

  • With cache: Falls back to stale cached data
  • Without cache (cold start): Server starts with 0 tools + status tool
  • The gameball-status tool always registers and shows the error state

First run: ~2-3 seconds (fetches from GitHub) Subsequent runs: ~200-500ms (validates cache with 1 API call)

Architecture

gameball-integrations/
ā”œā”€ā”€ src/
│   ā”œā”€ā”€ index.ts              # Entry point (stdio transport)
│   ā”œā”€ā”€ server.ts             # Server creation (parallel init, graceful failure)
│   ā”œā”€ā”€ status-tool.ts        # gameball-status tool
│   ā”œā”€ā”€ docs/                 # Documentation subsystem (7 tools)
│   │   ā”œā”€ā”€ index.ts          # Init + tool registration
│   │   ā”œā”€ā”€ fetcher.ts        # GitHub fetching + SHA-based cache
│   │   ā”œā”€ā”€ doc-indexer.ts    # MiniSearch indexing + chunking
│   │   ā”œā”€ā”€ mdx-parser.ts     # MDX → plain text
│   │   ā”œā”€ā”€ codegen.ts        # Multi-language code generation
│   │   ā”œā”€ā”€ types.ts          # Doc-specific types
│   │   └── tools/            # 7 documentation tools
│   ā”œā”€ā”€ integration/          # API integration subsystem (71 tools)
│   │   ā”œā”€ā”€ index.ts          # Client creation + tool registration
│   │   ā”œā”€ā”€ openapi-fetcher.ts # Independent openapi.json fetch + gzip cache
│   │   ā”œā”€ā”€ http-client.ts    # GameballClient (auth, error formatting)
│   │   └── generator.ts      # Auto-generates 71 tools from OpenAPI
│   └── shared/               # Common utilities
│       ā”œā”€ā”€ github-cache.ts   # GitHub fetch, SHA check, cache dir, timeout
│       ā”œā”€ā”€ types.ts          # Shared type definitions (ApiEndpoint, etc.)
│       └── openapi-parser.ts # OpenAPI JSON → ApiEndpoint map

Cache Layout

~/.cache/gameball-integrations/
ā”œā”€ā”€ manifest.json             # Docs cache manifest (commitSha, file list)
ā”œā”€ā”€ [__-delimited .mdx files] # Cached documentation pages
ā”œā”€ā”€ openapi-manifest.json     # OpenAPI cache manifest (commitSha, timestamp)
└── openapi.json.gz           # Gzip-compressed OpenAPI spec

Usage Examples

Learning about Gameball

Use docs-lookup-topic to learn about "customer referrals"
Use docs-get-doc to read "tutorials-new/referral-setup"
Use docs-generate-code-example for JavaScript customer creation

Making API Calls

Use create-customer to register a new customer
Use cashback-transaction to reward points
Use query-transactions to check transaction history

Checking Server Health

Use gameball-status to see subsystem status, config, and cache info

Development

# Install dependencies
npm install

# Run in development mode (with auto-reload)
npm run dev

# Build for production
npm run build

# Run built version
npm start

Requirements

  • Node.js >= 20
  • For integration tools: Valid Gameball API credentials

License

MIT

Links

Support

For issues or questions:

  • Documentation tools: Check cache at ~/.cache/gameball-integrations/
  • Integration tools: Verify API credentials in environment variables
  • GitHub Issues: [Report bugs or request features]

Recommended Servers

playwright-mcp

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.

Official
Featured
TypeScript
Magic Component Platform (MCP)

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.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

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.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

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.

Official
Featured
TypeScript
Kagi MCP Server

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.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

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.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured