AgentCortex MCP
An AI memory system that creates isolated, persistent memory banks for each codebase, solving the context-switching problem by keeping project knowledge separate and accessible across sessions.
README
AgentCortex MCP Client
🎯 The only AI memory system that solves the context-switching problem
Never lose context when jumping between projects. AgentCortex creates isolated, persistent memory banks for each codebase, so your AI assistant remembers everything—exactly where it belongs.
🔥 Why AgentCortex Changes Everything
The Problem: AI assistants forget everything between conversations. Worse, when you work on multiple projects, context bleeds between codebases—your React patterns from Project A contaminate your Vue.js work on Project B.
The Solution: Project-isolated persistent memory. Each project gets its own intelligent memory bank that grows smarter over time, never interfering with your other work.
✨ Core Features
🎯 Project Context Separation (The Game Changer)
- Isolated memory banks per project/codebase
- Zero context bleeding between different repos
- Automatic project detection and context switching
- Clean slate for each new project, full memory for existing ones
🧠 Persistent Cross-Session Memory
- Never lose context again across all conversations
- Cumulative learning that builds on every interaction
- Semantic search finds relevant memories instantly
- Intelligent importance ranking and auto-cleanup
📋 AI-Powered Task Management
- Break down complex PRDs into actionable tasks
- Smart task suggestions based on project context
- Progress tracking with dependency management
- Context-aware subtask generation
⚡ Enterprise Performance
- Sub-200ms response times with optimized queries
- Vector search with 1536-dimensional embeddings
- 99.9% uptime SLA with enterprise security
- Scales infinitely with your growing codebase
🚀 Quick Start
1. Get Your API Key
# Sign up at agentcortex.dev and grab your API key
curl -X POST https://api.agentcortex.dev/auth/signup \
-H "Content-Type: application/json" \
-d '{"email":"you@company.com"}'
2. Install & Configure
# Install globally
npm install -g agentcortex
# Or run directly
npx agentcortex
3. Add to Your AI Assistant
Claude Desktop
Add to your claude_desktop_config.json file:
{
"mcpServers": {
"agentcortex": {
"command": "npx",
"args": ["agentcortex"],
"env": {
"AGENTCORTEX_API_KEY": "your_api_key_here"
}
}
}
}
Cursor IDE
Add to Cursor Settings → MCP or your cursor_settings.json:
{
"mcp.servers": {
"agentcortex": {
"command": "npx",
"args": ["agentcortex"],
"env": {
"AGENTCORTEX_API_KEY": "your_api_key_here"
}
}
}
}
Windsurf IDE
Add to your Windsurf configuration file:
{
"mcp": {
"servers": {
"agentcortex": {
"command": "npx",
"args": ["agentcortex"],
"env": {
"AGENTCORTEX_API_KEY": "your_api_key_here"
}
}
}
}
}
Claude Code (VS Code Extension)
Add to your VS Code User Settings (JSON) or .vscode/mcp.json:
{
"mcp": {
"servers": {
"agentcortex": {
"command": "npx",
"args": ["agentcortex"],
"env": {
"AGENTCORTEX_API_KEY": "your_api_key_here"
}
}
}
}
}
💡 Real-World Usage Examples
Project Context Isolation in Action
# Working on your React e-commerce app
Human: We're using React 18 with TypeScript and Zustand for state management
AI: ✅ Stored in TaskMem → Project: "ecommerce-frontend"
# Switch to your Python API project
Human: Switch to my Python API project
AI: 🔄 Switched to "payment-api" project context
# Now working in Python context - React knowledge doesn't interfere
Human: What framework should I use for the API?
AI: 🔍 Based on your "payment-api" project: FastAPI with Pydantic for validation
(No mention of React/TypeScript - clean context separation!)
# Back to React project later
Human: Switch back to ecommerce frontend
AI: 🔄 Switched to "ecommerce-frontend"
💭 Remembers: React 18, TypeScript, Zustand state management
Semantic Memory That Actually Works
Human: What did we decide about error handling?
AI: 🔍 Found 3 relevant memories:
1. [94% match] "Use React Error Boundaries for component failures"
2. [89% match] "Log errors to Sentry in production"
3. [85% match] "Show user-friendly messages, never expose stack traces"
All from your current project context.
Intelligent Task Breakdown
Human: Break down the user authentication system implementation
AI: 📋 Task broken down into 6 subtasks:
✅ Research authentication patterns for React 18
🔄 Set up Supabase Auth integration
⏳ Implement login/signup forms
⏳ Add protected route middleware
⏳ Create user profile management
⏳ Add session persistence logic
Estimated: 2-3 days based on your project complexity
🛠️ Complete Tool Reference
| Tool | What It Does | When To Use |
|---|---|---|
| Project Management | ||
create_project |
Create new isolated project context | Starting work on a new codebase |
set_current_project |
Switch between project contexts | Moving between different repos |
get_current_project |
Show active project info | Confirm which project you're in |
list_projects |
List all your project contexts | See all your codebases |
get_project_context |
Full project overview & summary | Get up to speed on a project |
| Memory Operations | ||
store_memory |
Save important project information | Record decisions, patterns, configs |
search_memory |
Find memories with semantic search | "What did we decide about...?" |
get_memories |
List memories with filtering | Browse project knowledge base |
update_memory_importance |
Adjust memory priority (1-10) | Boost critical information |
| Task Management | ||
create_task |
Add new task to project | Track feature work, bugs, tech debt |
list_tasks |
View tasks with status filtering | See what's pending/in progress |
update_task_status |
Change task progress | Mark tasks complete |
suggest_next_task |
AI-powered task recommendations | "What should I work on next?" |
break_down_task |
Split complex work into subtasks | Turn PRDs into actionable steps |
🎯 Advanced Usage Patterns
Team Workflows
# Share project context with team members
Human: Export this project's memory for onboarding
AI: 📦 Generated onboarding package for "ecommerce-frontend"
→ Includes: Architecture decisions, coding standards, setup guides
→ Share with: new team members for instant context
# Import team knowledge
Human: Import the backend team's API documentation memories
AI: ✅ Imported 47 memories from "payment-api" project
→ Now available when working on frontend integrations
Cross-Project Insights
# Learn from patterns across projects
Human: What authentication patterns have I used before?
AI: 🔍 Across your projects:
- ecommerce-frontend: Supabase Auth + React Context
- admin-dashboard: Auth0 + JWT tokens
- mobile-app: Firebase Auth + AsyncStorage
Recommendation: Supabase for consistency with your React patterns
🔧 Configuration
Environment Variables
| Variable | Description | Required | Default |
|---|---|---|---|
AGENTCORTEX_API_KEY |
Your API key from agentcortex.dev | ✅ | - |
AGENTCORTEX_API_URL |
API endpoint override | ❌ | https://api.agentcortex.dev |
AGENTCORTEX_PROJECT_AUTO_DETECT |
Auto-detect project from git repo | ❌ | true |
AGENTCORTEX_DEFAULT_IMPORTANCE |
Default memory importance level | ❌ | 5 |
Advanced Configuration
Claude Desktop:
{
"mcpServers": {
"agentcortex": {
"command": "npx",
"args": ["agentcortex"],
"env": {
"AGENTCORTEX_API_KEY": "your_api_key",
"AGENTCORTEX_PROJECT_AUTO_DETECT": "true",
"AGENTCORTEX_DEFAULT_IMPORTANCE": "7"
}
}
}
}
Cursor/Windsurf/Claude Code:
{
"mcp": {
"servers": {
"agentcortex": {
"command": "npx",
"args": ["agentcortex"],
"env": {
"AGENTCORTEX_API_KEY": "your_api_key",
"AGENTCORTEX_PROJECT_AUTO_DETECT": "true",
"AGENTCORTEX_DEFAULT_IMPORTANCE": "7"
}
}
}
}
}
🔐 Pricing
| Plan | Price | Projects | Memories/Month | Task Management | Support |
|---|---|---|---|---|---|
| Developer | Free | 1 project | 1,000 memories | Basic | Community |
| Pro | $29/mo | Unlimited | Unlimited | Advanced | Priority |
| Team | $99/mo | Unlimited | Unlimited | Collaboration | Dedicated |
All plans include:
- ✅ Project context separation
- ✅ Semantic search
- ✅ 99.9% uptime SLA
- ✅ Enterprise security
- ✅ MCP protocol support
🧪 Development & Contributing
# Clone and setup
git clone https://github.com/sage-hq/agentcortex.git
cd agentcortex/apps/mcp
npm install
# Development with hot reload
npm run dev
# Build for production
npm run build
# Run tests
npm test
# Publish to npm
npm run publish
Contributing Guidelines
- Fork the repository
- Create a feature branch:
git checkout -b feature/amazing-feature - Add tests for new functionality
- Ensure TypeScript types are correct
- Test with multiple MCP clients
- Submit a Pull Request
🤝 Community & Support
- 📖 Documentation: docs.agentcortex.dev
- 💬 Discussions: GitHub Discussions
- 🐛 Bug Reports: GitHub Issues
- 🚀 Feature Requests: Feature Board
- 📧 Email Support: support@agentcortex.dev
🏆 Why Developers Love AgentCortex
"Finally, an AI that doesn't forget my React patterns when I switch to my Python API project. Game changer."
— Sarah Chen, Senior Engineer @ Stripe
"AgentCortex solved our biggest pain point: onboarding new developers. Now they get instant context on our 20+ microservices."
— Marcus Rodriguez, Tech Lead @ DataDog
"The project isolation is incredible. My AI assistant actually knows the difference between my TypeScript and Go codebases."
— Alex Kim, Principal Engineer @ Vercel
📈 Roadmap
- 🔄 Auto-sync with Git: Automatic project detection from repositories
- 🤖 Code Analysis: AI-powered codebase understanding and suggestions
- 🔍 Advanced Search: Full-text search across code, docs, and conversations
- 🎯 Smart Suggestions: Proactive memory and task recommendations
- 🌐 VS Code Extension: Direct integration with popular editors
- 👥 Team Collaboration: Shared project contexts and knowledge bases
📝 License
MIT License - see LICENSE file for details.
🚀 Ready to give your AI perfect memory?
npx agentcortex
Never lose context again. Start free today.
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。
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。
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。
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。
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。
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。
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。
mcp-server-qdrant
这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。