GPTaria MCP

io.github.gptaria-mcp由 GitHub 帳號 gptaria-mcp 發布v0.2.1更新於 Oct 24, 2026

Post a project brief & publish cases on GPTaria from Claude Code, Cursor or Windsurf.

已驗證STDIO僅桌面效率與工作流溝通與協作商業與電商

概覽

AI 產生的概覽

將 AI 助理連接到 GPTaria 平台，可發布專案需求、提交原型回應、發布案例並收發訊息。

可以做什麼
瀏覽與讀取專案、回應、澄清與回應時間軸，包括自己的回應及其狀態。查看訊息收件匣、會話時間軸和自己的公開個人檔案。發布專案需求、新增澄清、提交或更新回應，並完成報價、接受與拒絕的流程。發布作品集案例、產生案例 AI 摘要並編輯個人檔案。傳送訊息並將會話標示為已讀。查詢 AI 代幣餘額和價目表，取得存取通行證、AI 方案、秘密模式和案例驗證的結帳連結。
適用情境
適合在 GPTaria 上以客戶身分發布需求或以開發者身分提交原型回應的人，希望在 Claude Code、Cursor、Windsurf 或 Claude Desktop 中完成這些流程而不必開啟網站。正式使用前建議先以沙盒金鑰測試。
執行需求
透過 npm（npx gptaria-mcp）以 stdio 在本機執行，或作為 Claude Code 外掛安裝；說明中也提供託管連接器方式。需要 GPTaria 帳號，並在 GPTARIA_API_KEY 中提供於帳號設定建立的 API 金鑰。可用 GPTARIA_BASE_URL 覆寫平台基礎位址。需要連線至 GPTaria 的網路存取。
安裝前請注意
API 金鑰是僅顯示一次的機密，只應存放在 GPTARIA_API_KEY 中，測試時優先使用沙盒金鑰。許多工具會寫入或傳送資料：發布需求、回應、評論、案例、修改個人檔案和傳送訊息。多數操作會消耗 AI 代幣餘額，購買服務會回傳以格里夫納銀行卡付款的結帳連結，付款需由你確認。帳號角色與存取規則和網站一致。

安裝

在 SourceWeft 中

  1. 開啟 儀表板中的 GPTaria MCP，將其新增到工作區。
  2. 為需要使用其工具的對話啟用該服務。

Desktop only，透過 STDIO。 STDIO 服務會啟動本機處理程序，因此需要 SourceWeft 桌面主機。

其他 MCP 客戶端

參照 儲存庫 中的啟動說明。

README

GPTaria MCP

[npm] [CI] [MCP] [smithery] [Node] [License: MIT] [PRs Welcome]

GPTaria MCP connects your AI assistant — Claude Code, Claude Desktop, Cursor, Windsurf — to GPTaria and runs the whole flow straight from your editor: post a project brief, respond with a working prototype, message the other party, publish portfolio cases, and buy access — 23 tools. An MCP (Model Context Protocol) server for the GPTaria REST API.

  • Write in your own language — content is auto-translated into every platform language (the language is auto-detected from the text).
  • Prices in UAH (₴) — you pay by card with a Ukrainian bank card; list_services (free) lists exact prices.
  • Token billing — most operations spend a little of your AI-token balance; get_balance and list_services are free. You always confirm any payment.

GPTaria is an informational platform where clients post project briefs and builders answer with working prototypes. This repo is the developer surface: an MCP server that wraps the GPTaria REST API, plus ready-to-install Agent Skills that turn the tools into guided workflows.

REST API  →  MCP server  →  Agent skills(foundation)  (AI tools)     (guided workflows)

Tools

ToolForWhat it does
list_projectsanyoneList open projects (briefs) you can respond to
list_responsesclientsList a project's responses, to work on them in your own AI
list_my_responsesbuildersProjects you responded to, with status + win/offer flags
list_project_commentsanyoneList a project's client clarifications
get_response_threadbuildersYour own response's timeline (comments + versions)
list_conversationsanyoneYour message inbox (unread counts, last message)
get_conversationanyoneOne conversation's timeline (in your language)
get_balanceanyoneFree — AI-token balance + access status
list_servicesanyoneFree — the shop price list (in ₴)
create_projectclientsPublish a project brief via a Pro order
create_project_commentclientsAdd a public clarification to your own project
create_responsebuildersSubmit a response (proposal) to a project
post_proposal_updatebuildersPost an updated offer / comment in your thread
create_casebuildersPublish a portfolio case (draft by default)
generate_case_summarybuildersGenerate a case's AI summary block
update_profile / get_profileanyoneEdit / read your own public profile
offer_project / accept_project / decline_projectclient / builderThe offer→accept handshake
send_message / mark_conversation_readanyoneMessage the other party
buy_serviceanyoneFree tool — buy an access pass / AI pack / secret mode / case verification (returns a checkout link)

Every tool enforces the same rules as the website — account role, active access pass, the Super-Expert window, and payment where required. There is no bypass path; test against a sandbox key first.

Quick start

  1. Create an API key at gptaria.com/settings → API keys. The raw key is shown once — store it as GPTARIA_API_KEY.

    • Live keys: gptaria_live_…
    • Sandbox keys: gptaria_sbx_… — validate every call against the real rules but never persist ({ "dry_run": true }). Debug safely, then switch to live.

  2. Connect — pick one:

    • Hosted connector, no install (claude.ai / Claude Desktop → Settings → Connectors → Add custom connector): 
      https://www.gptaria.com/api/mcp?api_key=YOUR_KEY
    • Local, via npm — npx -y gptaria-mcp with GPTARIA_API_KEY in the server's env (Claude Code, Cursor, Windsurf).

    See docs/getting-started.md for the exact config per client, and for the full 23-tool reference.

  3. Install the skills (optional convenience layer) — copy the folders you want from skills/ into your assistant's skills directory.

No MCP? Every endpoint is a plain REST call too — see docs/api.md and examples/curl.md.

Install as a Claude Code plugin

In Claude Code, add this repo as a plugin marketplace and install it — the MCP server (from npm) and the skills come together:

/plugin marketplace add gptaria-mcp/gptaria-mcp/plugin install gptaria-mcp

Set GPTARIA_API_KEY in your environment first.

Contents

Install options

ChannelHow
npmnpx -y gptaria-mcp (or npm i -g gptaria-mcp) — see getting started
Claude Code plugin/plugin marketplace add gptaria-mcp/gptaria-mcp → /plugin install gptaria-mcp
From sourceclone, cd mcp-server && npm install && node index.mjs

Distribution manifests live in the repo: server.json (MCP registry), .claude-plugin/ (Claude Code), smithery.yaml (Smithery).

Links

License

MIT © GPTaria

來源：README.md，提交 5f50ada

工具

0
SourceWeft 尚未在沙箱中啟動過該軟體包，因此其工具未知。

版本歷史

1
  1. v0.2.1最新Sep 16, 2026

評分

已滿足適用檢查項的 50%

該評分衡量此收錄條目的完整程度，依據註冊中心條目、端點與公開來源計算得出。它不是對伺服器本身的評價，也不表示伺服器安全或目前可用：可連通性是帶日期的檢查結果，評分也不包含安裝量或使用者評價。

  • 近 30 天內可連通25 分

    請保持端點能夠回應 MCP 握手；SourceWeft 每 30 天探測一次。

  • 已宣告認證方式10 分

    請在註冊中心條目中宣告憑證請求標頭，或保持端點能夠回應探測，以便判定其認證方式。

  • 已列出工具15 分

    請讓伺服器在無憑證時回應 tools/list，或發布可在沙箱中無憑證啟動的軟體包。

  • 描述必需15 分
  • 圖示10 分
  • 文件15 分
  • 授權條款10 分