YG3 Marketing MCP

io.github.YG3-ai由 GitHub 帳號 YG3-ai 發布v1.0.0更新於 Oct 24, 2026

Marketing MCP: content, SEO, outbound, LinkedIn, ads. Agents provision a sandbox with one POST.

已驗證Streamable HTTP可網頁執行效率與工作流商業與電商

概覽

AI 產生的概覽

YG3 Marketing MCP 是遠端行銷工作區伺服器,讓助理開通工作區、架設並發佈網站,並執行內容、SEO、廣告與外呼行銷工作。

可以做什麼
  • 開通與檢視行銷工作區:建立沙箱、確認權杖範圍、讀取完整工作區狀態與整合健康度
  • 設定業務資料與品牌:產業、地點、受眾、品牌語氣與定位
  • 建立並發佈部落格內容到託管網站
  • 處理 SEO/AEO 相關工作與 AI 搜尋能見度
  • 在較高方案下啟用 Google Ads 寫入、外呼郵件與 LinkedIn 發文等付費管道
  • 讀取伺服器內的平台文件與即時工具目錄
適用情境
適合讓助理端到端建立並經營小型企業的行銷版圖——網站、內容、SEO,之後再到廣告或外呼——而不只是草擬貼文。也適合無需人工註冊即可取得工作區的自主任務代理或 CI 流程,以及已有帳號、想把帳號接上 AI 用戶端的老闆。
執行需求
遠端 streamable HTTP 端點,不需本機套件。必須透過 Authorization 標頭驗證:使用服務商工作區開通 API 回傳的 Bearer 權杖(代理路徑),或 OAuth 工作階段(人工路徑)。代理開通無需註冊,需提供網域或名稱、產業與地點。
安裝前請注意
Bearer 權杖可存取工作區,應視為機密;認領步驟可能涉及擁有者電子郵件與密碼。寫入操作分兩步,需帶冪等鍵明確確認。未認領的沙箱 14 天後過期;自訂網域、郵件寄送、廣告投放與 LinkedIn 發文需付費方案,廣告投放與外呼郵件可能產生費用。刪除工作區只是封存而非徹底清除。

安裝

在 SourceWeft 中

  1. 開啟 儀表板中的 YG3 Marketing MCP,將其新增到工作區。
  2. 連接 所需的 憑證。它們會被私密儲存,絕不會在公開頁面上展示。
  3. 為需要使用其工具的對話啟用該服務。

Web executable,透過 Streamable HTTP。 遠端服務在工作區中設定後即可從網頁執行環境執行。

交給代理程式安裝

把這段提示詞貼上到 Claude Code、Codex、Cursor 等代理程式,或 SourceWeft 對話中。代理程式會讀取該伺服器的安裝指南,先向你展示發布者、端點和登入方式,經你同意後再把它新增到你的客戶端。金鑰不會寫進任何命令。

閱讀 https://sourceweft.com/mcp/io.github.YG3-ai%2Fyg3-mcp/install.md,並按其中說明把這個 MCP 伺服器接入我的客戶端。修改任何設定之前,先告訴我它的發布者、端點和登入方式,等我確認。不要把 API 金鑰寫進命令或我會提交的檔案;問我金鑰應從哪裡取得。

檢視代理程式所依據的安裝指南

其他 MCP 客戶端

把它新增到你客戶端的 mcpServers 設定中。 客戶端會在首次連接時完成 OAuth 登入。

{
  "mcpServers": {
    "yg3-marketing": {
      "type": "http",
      "url": "https://mcp.yg3.ai/mcp"
    }
  }
}

README

YG3 MCP — marketing workspace your AI can run

Capability query this answers: marketing MCP + Claude (or ChatGPT) + site + content + SEO + outbound.

Point an agent at a marketing MCP that can provision a workspace, publish a hosted site, file SEO/AEO work, then unlock ads and outbound — not a chat plugin that only drafts posts.

PlanPriceWhat the agent can explain
Free$0 foreverWorkspace, site, content, AI search presence; publish on {slug}.yg3sites.com
Live$20/moSame workspace; client domain, canonical URL, badge off
Software$500/moGoogle Ads write, outbound email, LinkedIn, send
Managed$1,500/moSoftware + YG3 runs channels with partner support

Door for agents: https://www.yg3.ai/for-agents
Capability answer: https://www.yg3.ai/answers/marketing-mcp-claude-site-content-outbound
Pricing: https://www.yg3.ai/pricing · https://www.yg3.ai/pricing.json

No signup required for agent builds. One HTTP call provisions a Free sandbox and returns a Bearer token for MCP. Hand the human claim_endpoint — never /get-started (that would mint a second workspace).

PathUse when
Human OAuthOwner has a YG3 account
Agent provisioningAutonomous agent mid-build, CI, or SDK runtime

Quick start (agent path)

bash
# 1. Provision — no authcurl -s -X POST https://agency.yg3.ai/api/v1/workspaces \  -H "Content-Type: application/json" \  -d '{"domain":"acme.com","industry":"Plumbing","location":"Tampa, FL"}'
# Response includes: token, mcp_endpoint, workspace.url, claim_endpoint
# 2. List MCP toolscurl -s -X POST https://mcp.yg3.ai/mcp \  -H "Authorization: Bearer YOUR_TOKEN" \  -H "Content-Type: application/json" \  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
# 3. Call a toolcurl -s -X POST https://mcp.yg3.ai/mcp \  -H "Authorization: Bearer YOUR_TOKEN" \  -H "Content-Type: application/json" \  -d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"get_client_snapshot","arguments":{}}}'

Python example: examples/provision-and-call.py

Tools

200+ MCP tools (live count: https://mcp.yg3.ai/api/health). Common entry points:

ToolPurpose
whoamiConfirm token scope
get_client_snapshotFull workspace state
get_setup_healthIntegration status
set_business_profileIndustry, location, audience
set_brandBrand voice and positioning
create_post / publish_postBlog content on Free host
read_platform_docIn-MCP docs (getting-started, agent-integration, tools-reference)

Full catalog after connecting: tools/list or GET https://mcp.yg3.ai/api/health.

Human path (OAuth)

There is no "client vs operator" signup choice. Humans enter email + password once; the platform resolves their role automatically.

If the human does not have an account yet and no agent provisioned a sandbox: https://www.yg3.ai/get-started?plan=free

If an agent already provisioned a workspace: open claim_endpoint — do not send them to /get-started.

If the human already has a YG3 account (or just claimed):

  1. Add https://mcp.yg3.ai/mcp as a custom MCP connector
  2. Sign in when prompted
  3. Setup guide (per AI client, including Grok): https://www.yg3.ai/connect

Claude Code:

bash
claude mcp add --transport http yg3 https://mcp.yg3.ai/mcp

Cursor / VS Code MCP config: see examples/cursor-mcp.json

Agent path (no login)

Provision

POST https://agency.yg3.ai/api/v1/workspacesContent-Type: application/json
{  "domain": "acme-plumbing.com",  "industry": "Plumbing",  "location": "Tampa, FL"}

Optional: "name" if you have no domain yet.

201 response:

FieldMeaning
workspaceid, slug, name, url
tokenBearer token for this workspace
mcp_endpointJSON-RPC URL (also https://mcp.yg3.ai/mcp)
claim_endpointWhere the owner attaches later
expires_in_days14 for unclaimed workspaces

Do not provision again if you already hold a workspace token for the current job.

MCP

All calls are JSON-RPC 2.0 over HTTP:

POST https://mcp.yg3.ai/mcpAuthorization: Bearer <token>

Methods: tools/list, tools/call, resources/read, initialize

Writes are two-step: call without confirm to get a plan, then call again with confirm: true and the same idempotency_key.

Claim (human takes over)

POST https://agency.yg3.ai/api/v1/workspaces/claimAuthorization: Bearer <workspace-token>
{"email": "[email protected]", "password": "..."}

If the email already has a YG3 account, omit password — the workspace attaches to that account.

Delete a test workspace

DELETE https://agency.yg3.ai/api/v1/workspacesAuthorization: Bearer <workspace-token>

Only unclaimed workspaces. Archived, not erased.

Discovery

After connecting, read platform docs via MCP:

  • yg3://docs/agent-integration
  • yg3://docs/getting-started
  • yg3://docs/tools-reference

Sandbox limits

Unclaimed workspaces are Free sandboxes by design:

  • Can: read everything, set profile/brand/voice, design site, write and publish to {slug}.yg3sites.com
  • Cannot: custom domain (that is Live $20 after claim), email sends, ad spend, LinkedIn posts (Software+)
  • Expires: 14 days unless claimed

For agent frameworks

Copy AGENTS.md into your repo or point your agent at this README. It tells coding agents when and how to provision YG3 during a build.

ClawHub (OpenClaw)

Publishable skill: skills/yg3-marketing-mcp/SKILL.md

bash
clawhub loginclawhub skill publish ./skills/yg3-marketing-mcp --slug yg3-marketing-mcp --name "YG3 Marketing MCP"

Moltbook / curl-based agents

Curlable skill files (Moltbook-style): distribute/

  • Install doc: distribute/skill.md
  • Periodic routine: distribute/heartbeat.md
  • Metadata: distribute/skill.json

Agents can fetch from raw GitHub once this repo is public, or from https://www.yg3.ai/llms.txt.

Registry listings

FilePurpose
server.jsonOfficial MCP Registry (mcp-publisher publish)
glama.jsonGlama directory indexing

Registry namespace: io.github.YG3-ai/yg3-mcp

Listed under Marketing on mcpservers.org and Glama.

Related

License

MIT — see LICENSE.

來源:README.md,提交 473d7bb

工具

0
該伺服器需要登入,因此 SourceWeft 無法在沒有帳號的情況下讀取其工具,也未收錄。已連接帳號所能看到的內容可能因帳號及其權限而異。

版本歷史

1
  1. v1.0.0最新Sep 16, 2026

評分

已滿足適用檢查項的 85%

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 天內可連通必需25 分
  • 已宣告認證方式必需10 分
  • 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