MCP server exposing WhatsApp group chats to Claude — stdio + HTTP transports, chat intelligence processor, Cowork plugin
概览
An open-source MCP server that connects WhatsApp group chats to Claude, ChatGPT, Codex, and other MCP-compatible AI clients. It exposes eight tools for reading, searching, exporting, backfilling history, and replying to group conversations — plus a chat intelligence processor that extracts themes, opportunities, and actionable briefings from hundreds of messages. I built this because I'm in a dozen professional WhatsApp groups — practitioners, investors, founders — where the information density is remarkable and the retrieval rate is abysmal. WhatsApp is optimized for conversation, not comprehension. This server fixes that. give your AI client structured access to your WhatsApp groups: All tools support — say "Book Club" and it finds "Book Club — Monthly Reads" using Levenshtein distance + substring matching. Nobody remembers exact group names. The system shouldn't require you to.
README
以下为原文 README 节选，完整内容请查看原文。
WhatsApp MCP Server
Your WhatsApp groups are a live intelligence network. This server lets MCP-compatible AI clients read them.
An open-source MCP server that connects WhatsApp group chats to Claude, ChatGPT, Codex, and other MCP-compatible AI clients. It exposes eight tools for reading, searching, exporting, backfilling history, and replying to group conversations — plus a chat intelligence processor that extracts themes, opportunities, and actionable briefings from hundreds of messages.
I built this because I’m in a dozen professional WhatsApp groups — practitioners, investors, founders — where the information density is remarkable and the retrieval rate is abysmal. WhatsApp is optimized for conversation, not comprehension. This server fixes that.
What It Does
Eight MCP tools give your AI client structured access to your WhatsApp groups:
| Tool | What It Does |
|---|---|
whatsapp_list_groups |
Every group you belong to, sorted by recent activity |
whatsapp_get_messages |
Pull messages from any group with date range filtering |
whatsapp_group_info |
Metadata, participants, descriptions |
whatsapp_search_messages |
Keyword search across all groups or scoped to one |
whatsapp_export_chat |
Full export in WhatsApp’s native .txt format |
whatsapp_sync_history |
On-demand backfill of older group messages from WhatsApp servers |
whatsapp_send_message |
Send a message to any group (fuzzy name matching) |
whatsapp_reply_to_message |
Reply to a specific message as a quoted reply |
All tools support fuzzy group name matching — say “Book Club” and it finds “Book Club — Monthly Reads” using Levenshtein distance + substring matching. Nobody remembers exact group names. The system shouldn’t require you to.
A chat intelligence processor turns raw messages into structured briefings:
- Themes — recurring topics tagged as hot, important, or emerging
- Big ideas — intellectually interesting or actionable, flagged by relevance to your work
- Opportunities — collaboration, speaking, partnerships, content ideas, business leads
- Participation analysis — what you engage with vs. skip, where you’re visible vs. silent
- Live threads — active conversations worth jumping into, with suggested messages
- Notable quotes — standout lines worth saving or citing
- Cross-group synthesis — amplified signals, network nodes, compounding opportunities that only appear when you look across groups
安装
This server does not publish a one-line install command.
Open the repository installation guide 新手？先看 mcpskill.org 教程 →配置
{
"mcpServers": {
"whatsapp": {
"command": "node",
"args": ["/path/to/whatsapp-mcp-server/dist/mcp-server/index.js"],
"env": {
"WHATSAPP_SESSION_NAME": "my-session"
}
}
}
}