Memory Pickle MCP
Project management and session memory for AI agents that provides 13 tools for tracking projects, tasks, and context during AI coding sessions.
README
Memory Pickle MCP
Project management and session memory for AI agents. Provides 13 comprehensive MCP tools for tracking projects, tasks, and context during AI coding sessions.
suggestion
for kilo , cline , roo perform best with code mode your agent will not lose track. there are some struggle with HITL
Quick Start
Stable Version
{
"mcpServers": {
"memory-pickle": {
"command": "npx",
"args": ["-y", "@cabbages/memory-pickle-mcp"]
}
}
}
Compatibility
Extensively Tested & Optimized For:
- Cursor , Windsurf , Cline, Roo, Kilo Mode Code
Additional Support:
- Claude Desktop (direct MCP integration)
- VS Code with MCP extensions
Architecture Overview
Memory-Only Storage Architecture:
- All data exists only during the current chat session
- No files created on disk - zero filesystem dependencies
- Data automatically cleared when session ends
- Blazing fast performance with in-memory operations
- Complete session isolation for security
13-Tool Comprehensive System:
- 4 Read Tools: State recall, task listing, project overview, detailed task info
- 5 Write Tools: Project/task creation and updates, context switching
- 2 Memory Tools: Information storage and contextual search
- 2 Session Tools: Data export and handoff summary generation
Features
Hierarchical Task Organization
# Create parent task
create_task:
title: "Build Authentication System"
priority: "high"
# Create subtasks
create_task:
title: "Implement login API"
parent_id: "task_123" # Links to parent
create_task:
title: "Add password reset flow"
parent_id: "task_123"
Progress Tracking & Blockers
# Comprehensive task updates
update_task:
task_id: "task_123"
progress: 75
notes: ["API endpoints complete", "Working on validation"]
blockers: ["Waiting for security review", "Database schema pending"]
completed: false
Memory System with Importance
# Store critical decisions
remember_this:
content: "Switched from JWT to session cookies for security"
importance: "critical"
project_id: "proj_auth"
# Search by importance level
recall_context:
query: "security decision"
importance: "critical"
limit: 5
Workflow
Session Handoff
# Generate clean transition summary
generate_handoff_summary:
format: "detailed"
project_id: "current"
# Export for permanent storage
export_session:
format: "markdown"
include_handoff: true
raw_markdown: true # Clean output for processing
Integration Notes
- Context switching: Agents may lose context between tool calls (in some case)
- State synchronization: Use
recall_stateto refresh context - Error handling: Tools provide detailed error information send me issues <3
Requirements
- Node.js: 16.0.0 or higher
Links & Resources
- Official Website: pickle.cabbages.work
- GitHub Repository: Justar96/memory-pickle-mcp
- NPM Package: @cabbages/memory-pickle-mcp
- Pre-release: @cabbages-pre/memory-pickle-mcp-pre
Local Development
{
"mcpServers": {
"memory-pickle-dev": {
"command": "node",
"args": ["build/index.js"],
"cwd": "/path/to/your/memory-pickle-mcp"
}
}
}
Pre-release Version (Latest Development)
{
"mcpServers": {
"memory-pickle-pre": {
"command": "npx",
"args": ["-y", "@cabbages-pre/memory-pickle-mcp-pre@latest"]
}
}
}
Documentation
- Complete Tools Reference - All 13 tools with examples
- Usage Guide - Workflows and best practices
- Changelog - Version history and updates
Troubleshooting
Tools Not Working
- Update to latest version:
npx -y @cabbages/memory-pickle-mcpUse@latestflag - Verify MCP client configuration in settings
- Check Node.js version:
node --version - Restart MCP server for complete reset
Data Loss Prevention
- Tell Agent to use
generate_handoff_summarybefore ending sessions - Export critical data with
export_session - Save important decisions as markdown files
- Document project state in external files
Version Information
Current Version: 1.3.9
Recent Changes:
- 13 comprehensive tools for complete project management
- Advanced hierarchical task organization with parent-child relationships
- Code location linking for development context
- Importance-based memory classification system
- Comprehensive session export and handoff capabilities
- Optimized for modern AI coding assistants (Cursor, Cline, Roo, etc.)
License
Apache 2.0 - See LICENSE for details
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。