Memex Targeted Search Server
Enables AI agents to search through Memex conversation history and local project files to retrieve specific commands, code snippets, and technology overviews. It utilizes smart context management and faceted filtering to provide relevant search results without causing context overload.
README
Memex Targeted Search Server
A Model Context Protocol (MCP) server that provides targeted search capabilities across Memex conversation history and project files.
Overview
This MCP server enables AI agents to efficiently search through:
- Conversation History: 952+ conversation files from Memex with metadata, titles, summaries, and message content
- Project Files: 516+ project directories in the user's workspace with various file types and technologies
Features
🔍 Core Search Tools
search_conversations- Search conversation history by text, metadata, and filtersget_conversation_snippet- Retrieve specific parts of conversations without context overloadsearch_projects- Search project files by content, file types, and namesget_project_overview- Get project summaries with technology detectionfind_command- NEW! Find specific commands, CLI usage, or code snippets from conversation history
🎯 Smart Context Management
- Returns targeted snippets instead of full conversations
- Limits search scope to prevent context explosion
- Supports faceted filtering (dates, projects, file types)
- Provides relevance scoring for search results
Installation
# Clone the repository
git clone https://github.com/memextech/memex-targeted-search-server.git
cd memex-targeted-search-server
# Install dependencies
npm install
# Build the project
npm run build
Configuration
The server is configured to search:
- Conversation History:
~/Library/Application Support/Memex/history/ - Project Files:
~/Workspace/
MCP Server Configuration
Add to your MCP configuration (e.g., Claude Desktop config):
{
"mcpServers": {
"memex-search": {
"command": "node",
"args": ["/path/to/memex-targeted-search-server/dist/index.js"]
}
}
}
Usage Examples
1. Find Forgotten Commands
"I don't remember what the command is to run the memex agent cli"
find_command({
query: "memex agent cli",
command_type: "cli",
limit: 5
})
Find specific npm commands
find_command({
query: "npm install",
command_type: "cli",
limit: 5
})
Example Response:
{
"query": "npm install",
"total_found": 3,
"commands": [
{
"command": "npm install -g firebase-tools",
"context": "Install Firebase CLI: `npm install -g firebase-tools`\n- Login to Firebase: `firebase login`",
"conversation_id": "abc123",
"conversation_title": "Firebase Setup Guide",
"message_index": 7,
"confidence": 0.9,
"type": "cli"
}
]
}
2. Search Conversations
Find conversations about specific topics
search_conversations({
query: "3D modeling",
limit: 5
})
Example Response:
{
"total_found": 3,
"conversations": [
{
"conversation_id": "a3edfc8f-0978-415e-9de8-18f4d94ea3a2",
"title": "3D Interactive Solar System Model",
"summary": "Design an engaging, visually appealing 3D representation of planets and celestial bodies",
"created_at": "2025-05-27T17:13:26Z",
"project": "Stellar 3d solar system",
"message_count": 76,
"relevance": "content"
}
]
}
Filter by date range and project
search_conversations({
query: "python",
project: "cad_example",
date_from: "2025-01-01",
date_to: "2025-03-01",
limit: 3
})
3. Get Conversation Details
Retrieve specific messages from a conversation
get_conversation_snippet({
conversation_id: "bf283daa-25d3-434f-ad7e-9adda48cdcdd",
message_start: 1,
message_count: 3
})
Example Response:
{
"conversation_id": "bf283daa-25d3-434f-ad7e-9adda48cdcdd",
"title": "3D Model 3MF File Creation",
"message_range": "1-3",
"total_messages": 30,
"messages": [
{
"index": 1,
"role": "user",
"content": "can I create a 3D model in .3mf?"
},
{
"index": 2,
"role": "assistant",
"content": "I'll help you create a 3D model using PythonOCC and convert it to .3mf format..."
}
]
}
4. Search Projects
Find files by technology
search_projects({
query: "interface",
file_types: ["ts", "js"],
limit: 10
})
Search all project files
search_projects({
query: "streamlit",
limit: 5
})
Example Response:
{
"total_found": 3,
"results": [
{
"project": "ad_campaign_dashboard",
"file": "ad_campaign_dashboard/app.py",
"match": "import streamlit as st",
"line": 1
}
]
}
5. Get Project Overview
Analyze project structure and tech stack
get_project_overview({
project_name: "memex_targeted_search_server"
})
Example Response:
{
"name": "memex_targeted_search_server",
"path": "/Users/user/Workspace/memex_targeted_search_server",
"file_count": 8,
"directories": ["dist", "src"],
"file_types": {
"ts": 1,
"js": 1,
"json": 3,
"md": 1
},
"main_files": ["package.json", "README.md"],
"technologies": ["JavaScript/TypeScript"]
}
Real-World Usage Scenarios
Scenario 1: "I forgot that command..."
// User: "I don't remember what the command is to run the memex agent cli"
find_command({
query: "memex agent",
command_type: "cli",
limit: 5
})
// User: "What was that firebase command to deploy?"
find_command({
query: "firebase deploy",
command_type: "cli",
limit: 3
})
// Result: Finds exact commands with context from previous conversations
Scenario 2: Finding Related Work
// Agent: "I need to find previous conversations about Blender projects"
search_conversations({
query: "blender",
limit: 5
})
// Result: Finds 2 conversations about 3D Manhattan cityscape and geometric skyscraper
// Agent can then drill down into specific conversations for details
Scenario 3: Code Reference Lookup
// Agent: "Show me Python projects that use Streamlit"
search_projects({
query: "streamlit",
file_types: ["py"],
limit: 10
})
// Result: Finds specific Python files with Streamlit imports
// Agent can then examine project structure and implementation patterns
Scenario 4: Cross-Reference Discovery
// Agent: "Find conversations from January 2025 about 3D modeling"
search_conversations({
query: "3D model",
date_from: "2025-01-01",
date_to: "2025-01-31",
limit: 5
})
// Agent: "Now show me the related project files"
get_project_overview({
project_name: "cad_example"
})
API Reference
search_conversations
- Purpose: Search conversation history with flexible filtering
- Parameters:
query(required),limit,project,date_from,date_to - Returns: Array of conversation metadata with relevance scoring
get_conversation_snippet
- Purpose: Retrieve specific message ranges from conversations
- Parameters:
conversation_id(required),message_start,message_count - Returns: Conversation snippet with message details
search_projects
- Purpose: Search project files by content and metadata
- Parameters:
query(required),file_types,limit - Returns: Array of file matches with context
get_project_overview
- Purpose: Analyze project structure and technology stack
- Parameters:
project_name(required) - Returns: Project summary with file counts and tech detection
find_command
- Purpose: Find specific commands, CLI usage, or code snippets from conversation history
- Parameters:
query(required),command_type(cli/code/config/any),limit - Returns: Array of commands with context, confidence scoring, and conversation references
Architecture
Built with:
- TypeScript - Type-safe development
- MCP SDK - Official Model Context Protocol SDK
- Node.js - Runtime environment
- File System APIs - Direct file access for performance
Performance Considerations
- Limits search scope to prevent overwhelming results
- Uses streaming JSON parsing for large files
- Implements intelligent file filtering
- Caches frequently accessed metadata
- Returns truncated content with full context available on demand
Agent Experience
The server is designed for optimal agent interaction:
- Targeted Search: Find specific information without context overload
- Faceted Filtering: Multiple search dimensions (date, project, file type)
- Progressive Discovery: Start with summaries, drill down to details
- Context Preservation: Maintain conversation and project relationships
Development
Running in Development
npm run dev
Building for Production
npm run build
npm start
Testing
The server includes comprehensive error handling and graceful degradation for:
- Missing or corrupted conversation files
- Inaccessible project directories
- Invalid JSON parsing
- Large file handling
Contributing
Contributions are welcome! Please feel free to submit a Pull Request.
License
ISC License
🤖 Generated with Memex
Co-Authored-By: Memex noreply@memex.tech
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。