iMessage MCP
A Model Context Protocol server for reading iMessage data from macOS.
README
iMessage MCP
A Deno monorepo containing packages for iMessage access on macOS:
- @wyattjoh/imessage - Core library for read-only iMessage database access
- @wyattjoh/imessage-mcp - Model Context Protocol (MCP) server for LLM integration
Features
- Search messages by text content, contact, or date range
- Get recent messages
- List all chats/conversations
- Get all contacts/handles
- Retrieve messages from specific chats
- Search macOS Contacts by name with iMessage handle ID correlation
Requirements
- macOS (iMessage is only available on macOS)
- Deno 2.x or later
- Read access to
~/Library/Messages/chat.db - Read access to
~/Library/Application Support/AddressBook/(for contacts search)
Packages
@wyattjoh/imessage
Core library for accessing iMessage data:
deno add @wyattjoh/imessage
import { openMessagesDatabase, searchMessages } from "@wyattjoh/imessage";
const db = await openMessagesDatabase();
const results = await searchMessages(db, { query: "hello" });
db.close();
@wyattjoh/imessage-mcp
MCP server for LLM integration:
# Run directly from JSR
deno run --allow-read --allow-env --allow-sys --allow-ffi jsr:@wyattjoh/imessage-mcp
# Or install globally
deno install --global --allow-read --allow-env --allow-sys --allow-ffi -n imessage-mcp jsr:@wyattjoh/imessage-mcp
For Claude Desktop app integration, add this to your claude_desktop_config.json:
{
"mcpServers": {
"imessage": {
"command": "deno",
"args": [
"run",
"--allow-read",
"--allow-env",
"--allow-sys",
"--allow-ffi",
"jsr:@wyattjoh/imessage-mcp"
]
}
}
}
Option 2: From Source
- Clone this repository
- Install dependencies:
deno cache src/index.ts - Run the server:
deno run --allow-read --allow-env --allow-sys --allow-ffi src/index.ts # Or use the task: deno task start
Available Tools
-
search_messages - Search messages with filters
query(optional): Text to search forhandle(optional): Phone number or email to filter bystartDate(optional): ISO datetime string for start dateendDate(optional): ISO datetime string for end datelimit(optional): Maximum results (1-200, default: 100)offset(optional): Pagination offset (default: 0)
-
get_recent_messages - Get the most recent messages
limit(optional): Number of messages (1-100, default: 20)offset(optional): Pagination offset (default: 0)
-
get_chats - List all conversations
limit(optional): Number of chats (1-200, default: 50)offset(optional): Pagination offset (default: 0)
-
get_handles - Get all contacts/handles
limit(optional): Number of handles (1-200, default: 100)offset(optional): Pagination offset (default: 0)
-
get_messages_from_chat - Get messages from a specific chat
chatGuid(required): The chat GUIDlimit(optional): Number of messages (1-200, default: 50)offset(optional): Pagination offset (default: 0)
-
search_contacts - Search macOS Contacts by name and get phone numbers
firstName(required): First name to search for (e.g., 'John')lastName(optional): Last name to search for (e.g., 'Smith'). If omitted, searches across all name fieldslimit(optional): Maximum results (1-200, default: 50)offset(optional): Pagination offset (default: 0)- Returns contact info with phone numbers and email addresses that can be used as handle parameters
- Searches directly in the macOS AddressBook database for better performance and reliability
Pagination Examples
All tools now support pagination using limit and offset parameters and return pagination metadata:
// Get first 20 recent messages
get_recent_messages({ limit: 20, offset: 0 });
// Get next 20 recent messages (page 2)
get_recent_messages({ limit: 20, offset: 20 });
// Get first 10 chats
get_chats({ limit: 10, offset: 0 });
// Get messages 51-100 from a specific chat
get_messages_from_chat({
chatGuid: "iMessage;-;+15551234",
limit: 50,
offset: 50,
});
// Search with pagination
search_messages({
query: "meeting",
limit: 100,
offset: 200,
});
// Search contacts with pagination
search_contacts({
firstName: "John",
lastName: "Smith",
limit: 50,
offset: 0,
});
Response Format with Pagination Metadata
All paginated tools now return responses in this format:
{
"data": [
// Array of results (messages, chats, handles, etc.)
],
"pagination": {
"total": 1250, // Total number of results available
"limit": 100, // Current page size
"offset": 200, // Current offset
"hasMore": true, // Whether there are more results to fetch
"page": 3, // Current page number (1-indexed)
"totalPages": 13 // Total number of pages
}
}
This metadata helps you:
- Know the total number of results without fetching all of them
- Determine if there are more pages to fetch (
hasMore) - Calculate which page you're on and how many pages exist
- Build proper pagination UI components
Security Notes
- This server runs with read-only access to the iMessage database
- No messages can be sent or modified
- The server only accesses local data
Development
This is a Deno workspace monorepo. All commands run from the root affect all packages.
# Clone the repository
git clone https://github.com/wyattjoh/imessage-mcp.git
cd imessage-mcp
# Cache dependencies
deno cache packages/*/mod.ts
# Format all code
deno task fmt
# Lint all packages
deno task lint
# Type check all packages
deno task check
# Run tests
deno task test
# Run MCP server locally
cd packages/imessage-mcp
deno run --allow-read --allow-env --allow-sys --allow-ffi mod.ts
# Publish packages (CI/CD)
deno publish
Working on Individual Packages
# Work on @wyattjoh/imessage
cd packages/imessage
deno test --allow-read --allow-env --allow-ffi
# Work on @wyattjoh/imessage-mcp
cd packages/imessage-mcp
deno run --allow-read --allow-env --allow-sys --allow-ffi mod.ts
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。