WhatsApp MCP by TimelinesAI

io.github.InitechSoftware由 GitHub 账号 InitechSoftware 发布v1.0.2更新于 Oct 25, 2026

Ask questions across your WhatsApp inbox from Claude, ChatGPT, Cursor or any MCP client.

已验证Streamable HTTP可网页运行效率与工作流沟通与协作

概览

AI 生成的概览

一个托管式 MCP 服务器,把 AI 助手接入 TimelinesAI 的 WhatsApp 收件箱,以你的身份读取、回复、打标签和分配会话。

可以做什么
发现并读取会话与消息,可按状态、标签、负责人、电话和日期筛选,覆盖二维码连接的号码和 WhatsApp Business API 号码,并可查看单条消息详情和分页历史。发送消息、线程回复和表情回应,包括向新号码发起首次发送。整理会话:打开、关闭、分配或取消分配负责人,以及添加或移除标签。处理 WABA 模板:列出已批准的模板、查看其组件并发送,以及检查会话的服务窗口是否开启。查看工作区信息:套餐、席位、配额与计费周期、已连接的 WhatsApp 账号,以及团队成员的角色和邮箱。
适用场景
适合通过 TimelinesAI 处理 WhatsApp 客户沟通的团队,让助手从提示中总结哪些消息需要回复、起草并发送回复,或按标签和负责人整理会话。二维码连接的号码和 WhatsApp Business API 号码都适用。
运行要求
需要 TimelinesAI 账号,且工作区中至少连接一个 WhatsApp 号码;否则多数工具会返回空结果或报错。需要支持 Streamable HTTP 的远程 MCP 客户端,连接到托管端点,并在浏览器中通过 OAuth 登录一次。无需 API 密钥或安装;可选的插件方式需要 Node.js 18 及以上。
安装前请注意
它控制的是真实收件箱,没有沙箱:写入立即生效且无法撤销,收件人无法分辨消息来自助手。操作按你在 TimelinesAI 中的角色权限执行。发送会占用与界面相同的每月消息配额,通过 waba_start_chat 和 waba_chat_send_template 发送的 WABA 模板消息由 Meta 计费。批量操作前请先确认。

安装

在 SourceWeft 中

  1. 打开 控制台中的 WhatsApp MCP by TimelinesAI,将其添加到工作区。
  2. 连接 所需的 凭据。它们会被私密存储,绝不会在公开页面上展示。
  3. 为需要使用其工具的对话启用该服务。

Web executable,通过 Streamable HTTP。 远程服务在工作区中配置后即可从网页运行时运行。

交给智能体安装

把这段提示词粘贴到 Claude Code、Codex、Cursor 等智能体,或 SourceWeft 对话中。智能体会读取该服务器的安装指南,先向你展示发布者、端点和登录方式,经你同意后再把它添加到你的客户端。密钥不会写进任何命令。

阅读 https://sourceweft.com/mcp/io.github.InitechSoftware%2Fwhatsapp-mcp-by-timelinesai/install.md,并按其中说明把这个 MCP 服务器接入我的客户端。修改任何配置之前,先告诉我它的发布者、端点和登录方式,等我确认。不要把 API 密钥写进命令或我会提交的文件;问我密钥应从哪里获取。

查看智能体所依据的安装指南

其他 MCP 客户端

把它添加到你客户端的 mcpServers 配置中。 客户端会在首次连接时完成 OAuth 登录。

{
  "mcpServers": {
    "whatsapp-by-timelinesai": {
      "type": "http",
      "url": "https://mcp.services.timelines.ai/mcp"
    }
  }
}

README

WhatsApp MCP by TimelinesAI

Run your team's WhatsApp inbox from Claude, ChatGPT, Cursor or any AI assistant.

[License: MIT] [MCP] [Works with Claude] [Works with ChatGPT] [Works with Cursor] [smithery badge]

Connect your TimelinesAI WhatsApp inbox to Claude, ChatGPT, Cursor, Codex, Grok or any MCP client. Ask what needs a reply, then reply, react, label and assign chats from a prompt. 34 tools, all running as you, for both regular WhatsApp numbers and WhatsApp Business API (WABA) numbers.

Included in every TimelinesAI plan at no extra charge. Website · Setup guide · MCP Academy


What is this?

The TimelinesAI MCP server is a remote control for your real WhatsApp inbox inside TimelinesAI. Connect once with your TimelinesAI login, then your AI assistant gains 34 tools across chat discovery, messaging, triage (labels + assignments) and workspace introspection, for both regular WhatsApp numbers (connected by QR code) and WhatsApp Business API (WABA) numbers.

It's a hosted server: nothing to install, no API keys. Your assistant connects to https://mcp.services.timelines.ai/mcp and you sign in once.

No sandbox. Writes are real and visible to your contacts. Recipients cannot tell whether a message came from the UI or from an AI assistant: it's all just you.

Requirements

  • A TimelinesAI account with at least one connected WhatsApp number in your workspace. Without a connected number, most tools return empty results or errors. The MCP is a remote control for your real inbox, not a sandbox.
  • An AI assistant that supports remote MCP servers (all the clients below do).

Don't have a TimelinesAI account yet? Start a 10-day free trial →


Quick setup (~1 min)

Server URL for every client: https://mcp.services.timelines.ai/mcp

ChatGPT

Open the TimelinesAI plugin in ChatGPT, click Connect and sign in with your TimelinesAI account. Then ask: "@TimelinesAI what needs me today?"

Claude (claude.ai and Claude Desktop)

In Claude, open Customize → Connectors → Add custom connector, paste the server URL and sign in with your TimelinesAI account.

Claude Code

bash
claude mcp add --transport http timelinesai https://mcp.services.timelines.ai/mcp

Claude Code opens a browser tab the first time a tool is used. Sign in with your TimelinesAI account.

Prefer a plugin? /plugin marketplace add InitechSoftware/whatsapp-mcp-by-timelinesai, then /plugin install timelinesai-whatsapp@timelinesai. The plugin launches a local proxy, so it needs Node.js 18+; the command above doesn't.

Cursor

Add to Cursor (one click), or add this to ~/.cursor/mcp.json:

json
{  "mcpServers": {    "timelinesai": {      "url": "https://mcp.services.timelines.ai/mcp"    }  }}

The sign-in window opens on the first tool call.

Codex

Add to ~/.codex/config.toml:

toml
[mcp_servers.timelinesai]url = "https://mcp.services.timelines.ai/mcp"

Then run codex mcp login timelinesai once and approve in the browser. Codex doesn't start the sign-in on its own.

Grok Build

bash
grok mcp add timelinesai https://mcp.services.timelines.ai/mcp

Then type /mcps, select timelinesai and press i to sign in. Grok Build doesn't start the sign-in on its own.

Any other MCP client

Add https://mcp.services.timelines.ai/mcp as a remote server (Streamable HTTP) and sign in when prompted.

See examples/ for ready-to-copy config files.


First thing to try

After connecting, ask your assistant:

"What needs me today?"

or, for a tour:

"What tools does the timelinesai MCP server expose? Group them by purpose and give me one example prompt per group."


Tool catalog (34 tools)

Two parallel sets of tools (one for regular WhatsApp numbers connected by QR code, one for WhatsApp Business API (WABA) numbers) plus three workspace tools shared by both. Use the set that matches the number: workspace_whatsapp_accounts lists QR numbers, waba_accounts lists WABA numbers.

GroupTools
Workspace (shared)3
WhatsApp: QR-connected numbers14
WhatsApp Business API (WABA)17
Total34

Workspace: 3 tools

Read-only introspection. Cheap to call. Shared by both sets.

ToolPurpose
workspace_quotasCurrent plan, seats, messaging + API call quotas, billing period
workspace_whatsapp_accountsConnected QR WhatsApp accounts (id, phone, owner, status)
workspace_teamTeammates, roles, invitation status, account bindings. Use for assign emails in both sets

WhatsApp (QR-connected numbers): 14 tools

Chat discovery + inspection: 4 tools

Browse and read. No writes.

ToolPurpose
list_chatsFilter chats by status, labels, assignee, group/direct, phone, name, dates, WA account. Paginated 50/page
chat_detailsFull metadata for a single chat
chat_historyList messages in a chat. Filter by date and direction, page with message-UID cursors
message_detailsInspect a single message
Messaging writes: 4 tools

Quota-consuming. Same monthly messaging budget as the TimelinesAI UI.

ToolPurpose
chat_send_messageSend a message (text, attachment, or reply) in an existing chat
whatsapp_account_send_messageSend a message to any phone via a specific WA account (creates the chat if none exists: a cold send)
message_replyThreaded reply to a specific message
message_reactSet or clear an emoji reaction on a message
Chat mutations: 6 tools

State + triage operations on chats. Idempotent label and assign ops.

ToolPurpose
chat_openReopen a closed chat
chat_closeClose a chat
chat_assignAssign chat to a teammate by email (use workspace_team to discover emails)
chat_unassignUnassign current responsible teammate
chat_set_labelAdd a label to a chat
chat_remove_labelRemove a label from a chat

WhatsApp Business API (WABA): 17 tools

Accounts + chat discovery: 5 tools

Browse and read. No writes.

ToolPurpose
waba_accountsWABA numbers you can send from. Use id as waba_account_id
waba_list_chatsWABA chats you can see. Filter by open/closed, assignee, WABA account, phone; 50/page. Check service_window_is_open before sending
waba_chat_detailsOne WABA chat: service window, assignment, labels, WABA account
waba_chat_historyMessages in a WABA chat, newest first. Filter by date and direction, page with message-UID cursors
waba_message_detailsInspect one WABA message; failed sends include failure_reason
Templates: 2 tools
ToolPurpose
waba_templatesMessage templates you can use, 50/page. Only APPROVED templates can be sent
waba_template_detailsOne template's components, status, and the WABA accounts it can be sent from. Call before any template send
Messaging writes: 4 tools

Template sends incur a Meta charge. Free text only works while the chat's service window is open.

ToolPurpose
waba_start_chatStart a new chat with a phone number using an approved template. Meta charge
waba_chat_send_messageFree-text message in an existing WABA chat, only while service_window_is_open is true
waba_chat_send_templateSend an approved template in an existing WABA chat: the way to message when the service window is closed. Meta charge
waba_message_reactSet or clear an emoji reaction on a WABA message
Chat mutations: 6 tools

State + triage operations on WABA chats. After any mutation, waba_chat_details returns the authoritative state.

ToolPurpose
waba_chat_openReopen a closed WABA chat
waba_chat_closeClose a WABA chat
waba_chat_assignAssign a WABA chat to a teammate by email (from workspace_team)
waba_chat_unassignClear the WABA chat's assignee
waba_chat_set_labelAdd a label, keeping existing labels
waba_chat_remove_labelRemove one label, keeping the others; no-op if the label isn't on the chat

→ Full schemas and example prompts: docs/tools.md


How sign-in works

The server uses OAuth. The first time your assistant connects, a browser window opens on TimelinesAI. Sign in to your TimelinesAI account and approve access. Your assistant keeps the sign-in, so later sessions reconnect without asking.

If you're already signed in to TimelinesAI in your browser, approval can take a single click.

Every action runs as you: your role's permissions in TimelinesAI apply to MCP calls exactly as they apply to UI actions.


Safety + production behavior

⚠️ This MCP server controls a real WhatsApp inbox. Read this section before sending writes.

  • Messages sent via MCP are indistinguishable from messages sent in the UI. Recipients cannot tell.
  • Writes are real and immediate. No undo. Confirm before bulk actions.
  • Quota is shared with UI usage: MCP messages draw from the same monthly messaging budget. Call workspace_quotas to see headroom.
  • WABA template messages are charged by Meta. waba_start_chat and waba_chat_send_template send approved templates, which Meta bills for. Free-text waba_chat_send_message only works while the chat's service window is open.
  • No sandbox mode. Practice on your own number first if you're unsure.
  • Without a connected WhatsApp number, most tools return empty results or errors. Connect a number in TimelinesAI first.

Get started

  1. Start a 10-day free trial of TimelinesAI →
  2. Connect a WhatsApp number in the TimelinesAI dashboard.
  3. Pick your assistant above and connect it.
  4. Sign in, then ask: "What needs me today?"

Links


License

MIT © 2026 Initech Software / TimelinesAI

来源:README.md,提交 35d4f78

工具

0
该服务器需要登录,因此 SourceWeft 无法在没有账号的情况下读取其工具,也未收录。已连接账号所能看到的内容可能因账号及其权限而异。

版本历史

2
  1. v1.0.2最新Oct 8, 2026
  2. v1.0.1Sep 16, 2026

评分

已满足适用检查项的 85%

该评分衡量此收录条目的完整程度,依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价,也不表示服务器安全或当前可用:可连通性是带日期的检查结果,评分也不包含安装量或用户评价。

  • 近 30 天内可连通必需25 分
  • 已声明认证方式必需10 分
  • 已列出工具15 分

    请让服务器在无凭据时响应 tools/list,或发布可在沙箱中无凭据启动的软件包。

  • 描述必需15 分
  • 图标10 分
  • 文档15 分
  • 许可证10 分