Send4
ai.send4由 send4.ai 發布官方發布者v1.0.0更新於 Nov 4, 2026
Ask about and change your Send4 flows, campaigns, conversations, leads and calendar.
概覽
由 Send4 託管的 MCP 伺服器,讓 AI 助理讀取並修改你的 Send4 流程、活動、對話、潛在客戶與行事曆。
- 可以做什麼
- 讀取工作區資料:對話、潛在客戶、會議、成員、頻道、成效與訊息用量
- 檢視流程:列出與讀取流程、檢查流程、瀏覽卡片類型、範本與角色設定,查看執行紀錄、流失點與對話說明
- 建立流程:建立、複製與編輯草稿,儲存角色設定,建議卡片文案,演練流程,預覽或切換流程狀態
- 執行活動:預覽受眾與發送、建立草稿、發送、加入收件人、更新、暫停、恢復或取消,建立與標記潛在客戶
- 處理收件匣與行事曆:回覆、主動發起對話、指派或接手對話、恢復 AI、標記退訂、預約與更新會議
- 管理團隊與信箱:預覽並寄送成員邀請、變更角色、重寄邀請、查看信箱設定並建立信箱地址
- 適用情境
- 適合使用 Send4 訊息流程與活動的人:想用自然語言查詢成效、找出對話流失的環節、起草或修正流程、準備活動,或處理收件匣與行事曆事務,而不必在 Send4 應用程式裡逐項點選。
- 執行需求
- 遠端 Streamable HTTP 端點 mcp.send4.ai/mcp,無需安裝。需要 Send4 帳號且為付費方案,或已綁定信用卡的 14 天試用;需要 Client Admin 在工作區設定中開啟 Allow AI assistants;透過瀏覽器以 OAuth(含動態用戶端註冊)登入,每人使用自己的帳號。
安裝
在 SourceWeft 中
- 開啟 儀表板中的 Send4,將其新增到工作區。
- 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。
- 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/ai.send4%2Fmcp/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。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": {
"send4": {
"type": "http",
"url": "https://mcp.send4.ai/mcp"
}
}
}README
Send4 MCP server
Connect Claude, ChatGPT, Cursor, VS Code and other AI assistants to your Send4 workspace.
Ask about your flows, campaigns, conversations and usage in your own words, and let the assistant
make the changes your workspace allows.
[MCP: Streamable HTTP] [Sign-in: OAuth] [Hosted at mcp.send4.ai] [License: MIT]
Help centre · Tools · Permissions · Troubleshooting · Security
The Send4 MCP server is hosted by Send4 at https://mcp.send4.ai/mcp. There's nothing to
install or run: add the address to your assistant, sign in to Send4 in your browser, and choose
what the assistant may do. This repository holds the setup guides, the tool list and the place to
report a problem. It does not contain the server's code.
One-click setup
Pick your assistant. Cursor and VS Code install from the button; the others open a short guide.
Gemini CLI, Codex CLI and opencode: see Connect from the terminal.
Prefer to set it up by hand? Any assistant that supports remote MCP servers takes the address:
Contents
- What you can do
- Supported assistants
- Before you start
- How it works
- Tools and permissions
- Example workflows
- Data and security
- Admin notes: managing access
- Troubleshooting
- Security
- Support and feedback
What you can do
- Understand your flows: "What does my website form flow do?"
- Find out what went wrong: "Why did yesterday's conversation from the quote form stop?"
- See where people drop off: "Which flow lost the most people this week?"
- Check your numbers: "How did last week compare with the week before?" or "How close are we to our SMS limit this month?"
- Build and fix flows: "Build a draft flow that thanks a new lead, asks for their address and books a site visit."
- Run campaigns: "Create a draft campaign for leads tagged spring, and check it before launch."
- Work the inbox and calendar: "Which conversations are waiting for a person?" or "Find a free 30-minute slot on Thursday."
More in examples/prompts.md.
Supported assistants
Another assistant should work too if it connects to a remote MCP server over Streamable HTTP and signs in through your browser (OAuth with dynamic client registration).
Before you start
- A plan that includes it. Every paid plan does, and so does the 14-day trial once a card is on file. Starter has no campaigns, so campaign tools aren't offered there.
- A Client Admin switches it on. In Send4, open Settings → Integrations → AI assistants and turn on Allow AI assistants under Workspace access.
- You sign in as yourself. Each person connects with their own Send4 login.
How it works
- You add the address to your assistant.
- You sign in to Send4 in your browser. Send4 shows the account you're signed in to and the choices your role and your workspace allow. You tick what this assistant may do and select Connect.
- The assistant asks Send4 for its tools. It only hears about the tools this connection can use.
- You ask in your own words. The assistant picks the tools, and Send4 runs each one as you: your workspace, your role, your choices.
The workspace always comes from your sign-in. An assistant can't point a tool at another workspace.
Tools and permissions
Every connection can read. Each kind of change is a separate permission, off until a Client Admin allows it for the workspace and you tick it when you connect.
All tools
Open a group to see its tools. Each name links to that tool's full definition in tools/:
its description, its inputs and its safety hints, exactly as your assistant receives them.
Reading your workspace · 26 tools
For every role
For Client Admin and Client Agent
For Client Admin only
Reading customer messages · 1 tool
Building flows · 9 tools
Running campaigns · 12 tools
Working the inbox and calendar · 9 tools
Managing your team · 4 tools
Setting up email · 2 tools
Used by the in-chat views · 1 tool
- docs/tools.md has the same list, with which roles can use each tool.
- docs/permissions.md has the workspace switches and who can use what.
Example workflows
A Monday report
"Give me last week's report: conversations started, reply rate, meetings booked, how each running campaign did, and where people dropped off."
The assistant reads your performance, campaigns and drop-off, and writes the report. Nothing changes. Some assistants offer this as the ready-made weekly_report request.
Fix the card that loses people
"Which card in my quote flow loses the most people? Suggest clearer wording for it."
The assistant finds the card and leaves the new wording on it as a suggestion. You read it in the flow builder and accept it or not. Needs Build and change flows.
Change a live flow safely
"Make the opening message of my booking flow friendlier."
A live flow is never edited in place. The assistant duplicates it, changes the copy, rehearses it as a customer and checks it. Switching the copy on and the original off each wait for your yes. Needs Build and change flows.
Launch a campaign
"Count the leads tagged spring, create a draft campaign for them with the reactivation flow, sending on weekdays 9:00–17:00, and check it before launch."
The assistant counts the audience and creates the draft. Before launch, Send4 runs every check and works out the cost, and the assistant shows you that summary. Nothing is sent until you say yes. Needs Create and launch campaigns.
Hand a conversation to a person
"Which conversations are waiting for a person? Take over the one from this morning's form."
The assistant pauses the AI on that conversation so you can answer yourself. To hand it back, ask it to resume the AI. Needs Reply to customers and manage conversations.
Data and security
- You sign in with your Send4 account, through your browser. No API key or password goes into the assistant.
- It acts as you. Every tool runs in your workspace, with your role. It can't do more than you can in Send4.
- Your admin sets the limits, and you choose within them. The workspace switches decide what any assistant may do. When you connect, you tick what this one may do.
- Contact details are masked. Read answers mask phone numbers and email addresses. Customer messages are shared only when the workspace allows conversation content and you tick Read conversation messages.
- The big steps wait for your yes. These are launching a campaign, adding people to one, switching a flow on or a live flow off, inviting someone, and texting a number you typed. Before any of them, Send4 gives the assistant a summary to show you. After your yes, only what that summary described runs.
- A live flow is never edited in place. The assistant changes a copy, and you switch it over.
- No passwords in the chat. No tool takes a mailbox password. You connect your own mailbox yourself, in Send4.
- Every change is recorded. Changes appear in your Audit log, marked as made through an AI assistant.
- Requests are rate-limited per connection. A refused request says when to try again.
Where your data goes: what an assistant reads goes to the company behind it, under your own account with them. Check their terms before you allow conversation content.
Messages are data, not instructions: customers write the messages an assistant reads, and a message could try to give the assistant instructions. Don't let an assistant that reads conversations also manage your team.
Admin notes: managing access
Switch it on and choose what assistants may change
In Settings → Integrations → AI assistants, under Workspace access:
- Allow AI assistants lets people connect an assistant to their own account. On its own, an assistant can then read, and change nothing.
- Allow conversation content lets connections read messages and the details flows collected.
- Under What assistants may change, each kind of change has its own switch: Build and change flows, Reply to customers and manage conversations, Create and launch campaigns, Invite and manage team members and Create email addresses.
All of these are off to begin with. See docs/permissions.md.
Monitor what assistants change
Every change an assistant makes appears in Settings → Audit log, marked as made through an AI assistant, with the person who connected it.
Revoke access
- One person, one assistant: they open Settings → Integrations → AI assistants and select Disconnect beside it. Its access ends immediately.
- Everyone at once: a Client Admin turns off Allow AI assistants and selects Switch off. Every assistant in the workspace is disconnected, and nobody can connect one until an admin switches it on again.
Troubleshooting
More in docs/troubleshooting.md.
Security
Please report a security problem privately, never in a public issue. Use Report a vulnerability on this repository's Security tab, or see SECURITY.md.
Support and feedback
- A setup step is wrong or something doesn't work: open an issue. Issues are public: never paste access tokens, passwords, or your customers' messages, names, numbers or addresses.
- Something you'd like an assistant to do: open an issue with the Idea form.
- Account, plan or workspace questions: [email protected]
- Help centre: send4.ai/support/ai-assistant-tools
Licence
The guides and configuration in this repository are MIT licensed. Send4 itself is a commercial service, used under its terms.
來源:README.md,提交 28aa3e1
工具
0版本歷史
1- v1.0.0最新Nov 4, 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