AI Memory MCP
Persistent project memory MCP server for Claude Code that remembers development context across sessions, including features, fixes, notes, and standards.
README
AI Memory MCP - Project Local Memory Server
Project Overview
AI Memory MCP is an MCP server designed for Claude Code to provide persistent project memory. It enables AI to remember your project development journey across sessions, including feature planning, bug fixes, development notes, and code standards.
Key Features
| Feature | Description |
|---|---|
| Structured Memory | Categorizes project info into Features, Fixes, Notes, and Standards |
| Associations | Links between entries to build project knowledge graph |
| Tag System | Tag management for cross-dimensional queries and experience reuse |
| Three-Tier Architecture | Separated MCP Server, FastAPI Server, and Business Server |
| Async Design | Asynchronous architecture supporting concurrent access |
| Docker Deployment | Containerized deployment with persistent data storage |
System Architecture
The project uses a three-tier architecture with clear responsibilities:
┌─────────────────┐ ┌─────────────────┐
│ MCP Clients │ │ Web Clients │
│ (Claude Code) │ │ (Browser) │
└────────┬────────┘ └────────┬────────┘
│ │
▼ ▼
┌─────────────────┐ ┌─────────────────┐
│ MCP Server │ │ FastAPI Server │
│ (mcp_server) │ │ (rest_api) │
│ - SSE/HTTP │ │ - RESTful API │
└────────┬────────┘ └────────┬────────┘
│ │
└───────────┬───────────┘
▼
┌─────────────────────┐
│ Business Server │
│ (business) │
│ - Core business │
│ - Data storage │
│ - Tag system │
└─────────────────────┘
Directory Structure
ai_memory_mcp/
├── src/
│ ├── business/ # Business logic layer (core)
│ ├── mcp_server/ # MCP server layer
│ ├── rest_api/ # FastAPI REST API layer
│ ├── clients/ # Client modules
│ └── common/ # Common modules
├── docker/ # Docker deployment files
├── test/ # Test files (unit/integration/e2e/performance)
├── scripts/ # Utility scripts
├── examples/ # Example code (agents/skills)
├── docs/ # Documentation
├── config/ # Configuration files
├── scripts/ # Utility scripts
│ ├── start_mcp.py # MCP startup script
│ ├── start_business.py # Business service startup script
│ └── start_fastapi.py # FastAPI startup script
Quick Start
Method 1: Docker Deployment
# 1. Enter Docker directory
cd docker
# 2. Start service (default port 8000)
./manage.sh start
# 3. Check status
./manage.sh status
# 4. Stop service
./manage.sh stop
# 5. View logs
docker logs -f ai-memory-mcp
Method 2: Local Development
# 1. Create Conda environment
conda create -n ai_memory_mcp python=3.12
conda activate ai_memory_mcp
# 2. Install dependencies
pip install -r requirements.txt
# 3. Start business service
python scripts/start_business.py
# 4. Start MCP server (new terminal)
python scripts/start_mcp.py
# 5. Start FastAPI service (optional, new terminal)
python scripts/start_fastapi.py
Claude Code Configuration
Edit Claude Code configuration file and add MCP server:
{
"mcpServers": {
"memory": {
"url": "http://localhost:8000/mcp",
"type": "http"
}
}
}
Verify Installation
Test in Claude Code:
Please use MCP [memory] to list all projects
Core Features
Memory Content Categories
| Category | Purpose | Record Content |
|---|---|---|
| Features | Feature planning & tracking | Pending, in-progress, and completed features |
| Fixes | Bug fix records | Issue descriptions, severity levels, solutions |
| Notes | Development knowledge | Technical decisions, debugging processes, lessons learned |
| Standards | Project standards | Code conventions, naming conventions, architectural principles |
Memory Association System
┌─────────────────┐
│ Features │ ─┐
│ (Feature List)│ │ Associated
└─────────────────┘ │
├─→ Notes (Notes)
┌─────────────────┐ │ (Record development process/decision rationale)
│ Fixes │ ─┘
│ (Bug Fixes) │
└─────────────────┘
- Feature ↔ Note: Features can link to related notes (design decisions, reference docs)
- Fix ↔ Note: Bug fixes can link to troubleshooting processes
- Fix ↔ Feature: Bug fixes can link to specific feature modules
- Tag System: All entries support multi-tagging for cross-dimensional queries
MCP Tools List
Project Management
| Tool | Function |
|---|---|
project_register |
Register new project |
project_list |
List all projects |
project_get |
Get project info or query entries |
project_rename |
Rename project |
project_archive |
Archive project (MCP only) |
Entry Management
| Tool | Function |
|---|---|
project_add |
Add new entry (features/fixes/notes/standards) |
project_update |
Update entry content |
project_delete |
Delete entry |
Tag System
| Tool | Function |
|---|---|
tag_register |
Register new tag (define semantics) |
tag_update |
Update tag semantics |
tag_delete |
Delete tag |
tag_merge |
Merge tags |
project_tags_info |
Query tag information |
project_item_tag_manage |
Manage entry tags |
Query Features
- Filter by status (pending/in_progress/completed)
- Filter by severity (critical/high/medium/low)
- Filter by tags
- Filter by creation/update time
- Regex match on summary
- Paginated queries
Use Case Examples
Scenario 1: Continue Development Across Sessions
Context: Implemented API authentication middleware last week, continuing development today.
You: Continue improving API authentication, add refresh token mechanism
Claude: [Query memory] Found project records:
- feat_20260401: JWT authentication implemented using FastAPI Security
- note_20260401: Key stored at /secrets/jwt_key.pem
- standards_20260315: Auth-related functions need @auth_required decorator
Based on existing implementation, here's the refresh token design...
Scenario 2: Reuse Solved Experience
Context: New project encounters a Docker networking issue you solved before.
You: Services can't communicate after Docker Compose starts
Claude: [Cross-project query] Found similar issue in project web-dashboard:
- fix_20260310: Docker network communication failure
- Solution: Use custom network network_mode: bridge
- note_20260310: Need to explicitly declare networks in docker-compose.yml
Generated same configuration for new project...
Scenario 3: Follow Project Conventions
Context: Make AI generate code following team standards.
You: Implement user order query endpoint
Claude: [Query standards] According to project standards:
- All endpoints return wrapped Result<T> format
- Pagination uses PageRequest(page, size)
- Async functions must include _async suffix
Generated code:
async def get_orders_async(req: PageRequest) -> Result[List[Order]]:
...
Data Storage
Storage Location
Docker Mode: Persistent volume inside container at /var/lib/project_memory_ai/
Local Mode: data/ directory in project root
Data Directory Structure
/var/lib/project_memory_ai/
├── _metadata.json # Global metadata (project list, statistics)
├── _stats.json # API call statistics
│
├── project_a/ # ProjectA directory (project name)
│ ├── _project.json # Project metadata (id, name, info, _version)
│ ├── _tags.json # Tag registry (tag -> description, aliases)
│ ├── _group_configs.json # Group configurations (optional)
│ │
│ ├── features/ # Features group
│ │ ├── _index.json # Feature index (id -> summary, status, tags...)
│ │ ├── feat_20260408_001.json # Feature details
│ │ └── feat_20260408_002.json
│ │
│ ├── fixes/ # Bug fixes group (same structure as features)
│ ├── notes/ # Notes group (same structure as features)
│ └── standards/ # Standards group (same structure as features)
│
├── project_b/ # ProjectB directory
│
└── .archived/ # Archived projects
├── 20260408_123456_project_a.tar.gz
└── 20260408_123456_project_a.meta.json
Development Guide
Requirements
- Python 3.12+
- Conda
- Docker (for containerized deployment)
Running Tests
# Use test script
./scripts/run_tests.sh
# Or manually run
pytest test/ -v --cov=src/business --cov-report=html
Code Standards
- Use
blackfor code formatting - Use
rufffor code linting - Use
mypyfor type checking
Project Status
| Project | Version | Dev Branch | Main Branch |
|---|---|---|---|
| ai_memory_mcp | v1.0.0 | dev | main |
Development Statistics
- Features: 63
- Notes: 125
- Fixes: 24
- Standards: 18
- Tags: 59
Contributing
Contributions welcome! Please see CONTRIBUTING.md for details.
License
This project is licensed under the MIT License. See LICENSE file for details.
Related Links
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.
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.
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.
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.