WhatsApp MCP Server (Docker Edition)
Enables AI clients to interact with WhatsApp via MCP tools for searching contacts, sending/receiving messages, and managing chats, running in a Docker container with HTTP/SSE transport.
README
WhatsApp MCP Server (Docker Edition)
A WhatsApp MCP (Model Context Protocol) server that runs as a Docker container and is accessible over HTTP/SSE. Based on whatsapp-mcp-ts with the transport layer changed from stdio to SSE for remote access.
Features
- 8 MCP Tools: search_contacts, list_messages, list_recent_messages, list_chats, get_chat, get_message_context, send_message, search_messages
- Docker-first: Runs as a container, accessible via IP
- Streamable HTTP Transport: stateless MCP over HTTP POST
- Bearer-token auth for
/sse(setMCP_AUTH_TOKEN) - Persistent Data: SQLite DB (WAL + FTS5 search) and WhatsApp auth in Docker volumes
- Health Check:
/healthreturns 503 when WhatsApp is disconnected - Multi-Arch: Builds for both amd64 and arm64
Quick Start
1. Configure secrets
cp .env.example .env
# Generate a strong token and write it to .env:
echo "MCP_AUTH_TOKEN=$(openssl rand -hex 32)" >> .env
2. Pull and Run
# Pull the image
docker pull ghcr.io/YOUR-USERNAME/whatsapp-mcp-docker:latest
# Run with docker compose
docker compose up -d
# Watch logs for QR code
docker logs -f whatsapp-mcp
3. Scan QR Code
On first start, a QR code URL will appear in the logs. Open it in your browser and scan with WhatsApp (Settings → Linked Devices → Link a Device).
4. Configure MCP Client
This server speaks Streamable HTTP MCP (POST /sse), not legacy SSE. The exact config key depends on your client:
Gemini CLI (~/.gemini/settings.json — the key is httpUrl, not url):
{
"mcpServers": {
"whatsapp": {
"httpUrl": "http://192.168.0.101:3010/sse",
"headers": {
"Authorization": "Bearer YOUR_TOKEN_FROM_DOTENV"
}
}
}
}
⚠️ Common pitfall: Using
"url"in Gemini CLI selects the legacy SSE transport, which this server does not implement. You will get 401 errors that look like an auth problem but are actually a transport mismatch. Use"httpUrl".
Claude Desktop / Cline / Continue / Cursor (most use url for Streamable HTTP):
{
"mcpServers": {
"whatsapp": {
"url": "http://192.168.0.101:3010/sse",
"headers": {
"Authorization": "Bearer YOUR_TOKEN_FROM_DOTENV"
}
}
}
}
Contact Name Resolution
WhatsApp Web (Baileys) does not give us your phone-side address book directly. To still show your saved names — not the contact's self-chosen push names — the server uses three layers, in this order:
- Manual override (
contact_overrides.json, see below) — wins everything. - Auto-linking — every incoming message's
pushNameis captured and stored as thenotifyof the corresponding@s.whatsapp.netcontact. This lets the existing fuzzy-match in queries automatically connect the@lidform (used in groups & status updates) to the@s.whatsapp.netrow that holds your saved address-book name. After enough message activity, most contacts resolve themselves. - Fallback —
name→notify→ phone digits → JID.
Manual overrides (contact_overrides.json)
Use this only for contacts that auto-linking can't fix (e.g. a contact who has never written you directly, or whose push name and saved name are completely different and the fuzzy match doesn't catch it).
Path: <data dir>/contact_overrides.json (typically /opt/docker/whatsapp-mcp/data/contact_overrides.json on the host)
{
"210444891463794@lid": "Tammy",
"11390320418995@lid": "Amelie",
"4917697335710@s.whatsapp.net": "Tammy"
}
- The override wins over the DB-resolved name in every MCP response (
chat_name,sender_name,last_sender_name). - File is hot-reloaded on save — no container restart required.
- Each entry is a JID → display name mapping. Find a contact's JID via
search_contactsorlist_chats. - See
contact_overrides.example.jsonin the repo for the format.
Security
- Always set
MCP_AUTH_TOKENunless you bind the container to127.0.0.1only. Without a token, anyone on your network can read your chats and send messages from your account. - Without TLS, the token is sent in plaintext. For LAN-only deployments this is usually acceptable; for anything wider, terminate TLS in a reverse proxy (Caddy, Traefik, nginx).
- Logs in
/app/data/wa-logs.txtand/app/data/mcp-logs.txtmay contain message snippets. Treat the data volume as sensitive. - The QR code from initial pairing is logged at
infolevel once — it grants full session access for a few seconds. Don’t share that log.
Endpoints
| Endpoint | Method | Auth | Description |
|---|---|---|---|
/sse |
POST | Bearer | Streamable HTTP endpoint for MCP (stateless) |
/health |
GET | none | 200 when WhatsApp connected, 503 otherwise |
MCP Tools
| Tool | Description |
|---|---|
search_contacts |
Search contacts by name or phone number |
list_messages |
Get message history for a chat (paginated) |
list_chats |
List chats sorted by activity or name |
get_chat |
Get details of a specific chat |
get_message_context |
Get messages around a specific message |
list_recent_messages |
Cross-chat time-window query (e.g. "last 8 hours") with optional ISO since/until |
send_message |
Send a text message to a user or group |
search_messages |
Full-text search across messages |
Environment Variables
| Variable | Default | Description |
|---|---|---|
MCP_PORT |
3010 |
HTTP server port |
MCP_AUTH_TOKEN |
(unset) | Bearer token required on /sse. If unset, endpoint is open. |
LOG_LEVEL |
info |
Log level (debug, info, warn, error) |
WHATSAPP_MCP_DATA_DIR |
/app/data |
SQLite database directory |
WHATSAPP_AUTH_DIR |
/app/auth_info |
WhatsApp auth credentials directory |
TZ |
Europe/Berlin |
Timezone used for human-readable timestamps |
WHATSAPP_DEBUG |
(unset) | If true, prints @lid mapping diagnostics on startup |
Development
Requires Node.js >= 24 (stable node:sqlite and on-by-default TypeScript stripping). On Node 22/23, run with --experimental-sqlite --experimental-strip-types.
# Install dependencies
npm install
# Run locally (Node 24+)
node src/main.ts
# Run locally (Node 22/23)
node --experimental-sqlite --experimental-strip-types src/main.ts
# Type-check only
npx tsc --noEmit
# Build Docker image locally
docker build -t whatsapp-mcp-docker .
Architecture
AI Client (Gemini CLI, Claude, etc.)
│
│ HTTP POST (Streamable)
▼
┌─────────────────────┐
│ Express Server │ Port 3010
│ ├── POST /sse │ Stateless MCP
│ └── GET /health │ Health check
├─────────────────────┤
│ MCP Server │ 7 Tools
├─────────────────────┤
│ Baileys │ WhatsApp Web API (Reconnects + Backoff)
├─────────────────────┤
│ SQLite │ Messages & Contacts (Batch Sync)
└─────────────────────┘
Credits
Based on whatsapp-mcp-ts by jlucaso1.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.