Authenticated, self-hosted WhatsApp MCP server that serves 96 tools for the token cost of 29, with a labeled eval for tool-retrieval accuracy. Go bridge + Python FastMCP.
A self-hosted, authenticated MCP server for WhatsApp: 96 tools over your own message history, with server-side tool retrieval so the model pays for 29 of them instead of 96, and a labeled eval that measures whether retrieval actually picks the right tool.
Runs as two containers on a small VPS. Your messages stay on your box; the only credential is a bearer token you generate.
The reference WhatsApp MCP servers are stdio-only, unauthenticated, and expose a flat tool list. That is fine on a laptop and unworkable as an always-on service. Three problems show up immediately, and most of this repo is the answer to them.
1. Tool definitions are a fixed tax. MCP injects every tool definition into the model's context on every request, whether or not the request touches WhatsApp. At 96 tools that is roughly 20k tokens you pay for asking "what time is it". Dropping tools to save tokens means losing capability.
2. The same person has two identities. WhatsApp hides phone numbers behind LIDs inside groups, so one human appears under their phone number in DMs and under an opaque LID in group history. Any tool that filters on a single sender ID silently returns half their messages, with no error to tell you.
3. The app's UI is not the protocol. "View once", "delete for everyone", poll selection caps: these are conventions the client draws, not guarantees the server enforces. A server that reads the wire will disagree with the app, and when it does, the app is usually the one that is wrong.
From the project README.
Add the radar badge to your README — it shows your project was picked up by MCP Radar and links to this page:
[](https://mcp.liqiwa.com/s/HalemoGPA--whatsapp-mcp-server.html)
Production-ready MCP server that lets AI agents build, schedule, run and recover Apache DolphinScheduler data pipelines - 53+ tools, beyond read-only.
AdamRussak/self-docSelf-hosted documentation RAG pipeline that serves semantic search to LLM agents over the Model Context Protocol.
kenb38291-tech/fastmcp-sqliteProduction-grade, token-optimized FastMCP SQLite Server with sub-millisecond O(1) schema discovery, opcode execution watchdog, and fuzzy self-healing.
JoseJimenez-M/cortex-mcpSelf-hosted MCP server that lets Claude, ChatGPT, Meta Muse and any MCP client read and write one Markdown vault (Obsidian-compatible). Go, OAuth 2.1 with passkeys, no shell, no hard deletes.
Stupidoodle/swissdevjobs-cliSearch & apply to ~4,700 salary-transparent tech jobs across 7 countries (🇨🇭🇩🇪🇬🇧🇺🇸🇨🇦🇳🇱🇫🇷) from your terminal or AI agent — zero-dependency Python CLI + MCP server + Claude Code plugin
bybit-exchange/kaasTurn scattered notes, docs and transcripts into a queryable Markdown wiki — an LLM knowledge-base compiler with MCP access, no embeddings, self-hosted.
The top new MCP servers of the week, every Monday. No spam, unsubscribe anytime.