QUPU TECH · MCP SERVICE · EST. 2026

Make AI Agents Understand quputech.com

The QupuTech MCP service: six atomic tools that let in-browser AI agents read our games, jobs, and site content — and submit partnership inquiries. Built on the Model Context Protocol (standardized by Anthropic in November 2024) and deployed on Cloudflare Workers.

https://quputech.com/mcp LIVE · Streamable HTTP / JSON-RPC 2.0

The "USB-C" interface for AI ↔ services

Model Context Protocol (MCP) is an open standard (Anthropic, November 2024; current revision 2025-06-18) that lets LLMs exchange structured data with local or remote services. Claude Desktop, Cursor, Continue, Cline, and Zed all ship native MCP support, and Cloudflare's edge auto-injects a WebMCP bridge.js on quputech.com.

The QupuTech MCP endpoint wraps the public site — games, jobs, business plan, software copyrights — into six atomic tools. One JSON-RPC call returns structured data: no HTML scraping, no OCR, no URL plumbing.

STEP 01 Browser AI Agent Chrome 146+ · Claude / Cursor
STEP 02 POST /mcp JSON-RPC 2.0 · Streamable HTTP
STEP 03 Worker (DO) Zod schema · 60s cache
STEP 04 Pages JSON /pages-data/*.json

Every tool ships its own JSON Schema

Click the JSON badge in the top-right corner of a card to fetch the full input/output schema and feed it straight into your agent pipeline.

01

list_games

JSON

List every QupuTech title. Filterable by language and release status. Returns slug, name, tagline, status, release date, version, category, site URL, YouTube ID, CTA, and description.

localeenum[zh|en|zh-tw] statusenum[released|in_development|all]
Request / Response
JSON-RPC · tools/call
{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"list_games","arguments":{"locale":"en","status":"released"}}}
02

get_game

JSON

Fetch the complete record for one game by slug — localized names, icon, hero image, YouTube URL, external links. Use it for "Tell me about Cyber Cultivation 2066" queries.

slugenum[eggtopia|cyber|towerbound|pokid|mystery] localeenum[zh|en|zh-tw]
Request / Response
JSON-RPC · tools/call
{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"get_game","arguments":{"slug":"cyber","locale":"en"}}}
03

list_jobs

JSON

List every open position at QupuTech — title, location, category, responsibilities, requirements, application email, and anchor URL. Built for "Are you hiring?" flows.

localeenum[zh|en|zh-tw] include_nice_to_havebool
Request / Response
JSON-RPC · tools/call
{"jsonrpc":"2.0","id":4,"method":"tools/call","params":{"name":"list_jobs","arguments":{"locale":"en","include_nice_to_have":true}}}
04

search_site

JSON

Full-text search across the QupuTech site. Covers the entire sitemap, games, jobs, and publications. Case-insensitive substring matching, ranked by relevance.

querystring · 1-50 localeenum[all|zh|en|zh-tw] limitint · 1-20
Request / Response
JSON-RPC · tools/call
{"jsonrpc":"2.0","id":5,"method":"tools/call","params":{"name":"search_site","arguments":{"query":"towerbound","limit":8}}}
05

navigate

JSON

Return canonical URLs for key quputech.com pages. Designed for browser navigation or for showing users a direct link. Supports home / games / careers / business plan / privacy / terms / contact / MCP endpoint.

pageenum[home|games|careers|business_plan|privacy|terms|contact|mcp_endpoint] localeenum[zh|en|zh-tw|auto]
Request / Response
JSON-RPC · tools/call
{"jsonrpc":"2.0","id":6,"method":"tools/call","params":{"name":"navigate","arguments":{"page":"careers","locale":"en"}}}
06

submit_contact

JSON

Submit the partnership contact form. During the demo phase the lead lands in Cloudflare KV (LEADS) and a structured log is emitted to the Worker console. Production can switch to Resend / MailChannels for direct email delivery.

namestring · 1-80 emailemail topicenum[cooperation|investment|press|careers|other] messagestring · 10-2000
Request / Response
JSON-RPC · tools/call
{"jsonrpc":"2.0","id":7,"method":"tools/call","params":{"name":"submit_contact","arguments":{"name":"Alex","email":"alex@example.com","topic":"cooperation","message":"Interested in publishing collaboration","locale":"en"}}}

Paste this snippet into your MCP config

Claude Desktop uses ~/.claude/mcp.json; Cursor / Continue use .mcp.json at the project root. Restart the client and six quputech tools will appear in the tool palette.

JSON · .mcp.json
{
  "mcpServers": {
    "quputech": {
      "type": "http",
      "url": "https://quputech.com/mcp"
    }
  }
}

1Save the snippet above as ~/.claude/mcp.json (Claude Desktop) or project-root .mcp.json (Cursor / Continue).

2Restart the client → six quputech tools appear in the tool palette, marked with the QUPU TECH badge.

3Ask "List QupuTech's released games." The agent automatically calls list_games({status:"released"}) and renders the three titles in a structured response.

An open endpoint, optimised for LLMs

Built on the industry-standard Model Context Protocol. Drop-in support for Claude Desktop, Cursor, Continue, and other mainstream AI clients.

ProtocolModel Context Protocol, JSON-RPC 2.0
TransportStreamable HTTP (POST / JSON, SSE / EventStream)
Endpointhttps://quputech.com/mcp
Version2025-06-18 (backwards-compatible with 2025-03-26)
Compatible clientsClaude Desktop · Cursor · Continue · Cline · Zed · Windsurf · in-browser AI agents

Pipe quputech.com into your AI stack

Support / Community

Discord / Lark / WeCom Bots

Customer-support bots call list_games, get_game, and navigate directly. "Does QUPU have new games?" returns structured answers without maintaining a mirror database.

Hiring / ATS

Candidate Assistants

Recruiting platforms plug into list_jobs. Candidates asking "Any Flutter engineer roles?" immediately get responsibilities, apply_email, and the anchor_url for the careers page.

Partnerships

Proactive Visitor Conversations

Via Cloudflare's bridge.js, in-browser AI agents can call submit_contact directly. A visitor saying "I'd like to talk to your BD team" submits the lead straight into the LEADS KV namespace.

The five things people ask first

What is MCP and how is it different from Function Calling?

MCP is an open protocol standardized by Anthropic in November 2024. It defines how an LLM talks to a service — transport, session management, tool description — so any MCP-compatible client can consume any MCP endpoint without bespoke integration. Function Calling is each LLM vendor's proprietary tool-use mechanism (OpenAI, Anthropic, Google each ship their own); MCP is a vendor-neutral protocol layer that sits underneath.

Will my MCP client work out of the box? Do I need an SDK?

No SDK required. Every mainstream MCP client already speaks Streamable HTTP transport — just add { "type": "http", "url": "https://quputech.com/mcp" } to your config. If you want to build a custom client, the official SDK lives at @modelcontextprotocol/sdk on npm.

How will the protocol version evolve? When is 2025-11 shipping?

The current service targets MCP 2025-06-18 and stays backwards-compatible with 2025-03-26. When Anthropic ships a new revision, we handle the upgrade internally and keep a compatibility window for existing clients. We'll announce breaking changes in the changelog 30 days ahead.

Five lines of JSON, full agent access

Fully public during the demo phase — no auth and no billing.
For commercial usage, integration help, or protocol feedback, drop us a line.