@ai-devkit/memory

A lightweight MCP-based memory service for AI agents. Store and retrieve actionable knowledge using SQLite with FTS5 full-text search.

Features

• 🔍 Full-Text Search - FTS5 with BM25 ranking
• 🏷️ Tag-Based Filtering - Boost results by contextTags
• 📁 Scoped Knowledge - global, project, or repo-specific rules
• 🔄 Deduplication - Prevents duplicate content
• ⚡ Fast - SQLite with WAL mode, <50ms search latency
• 📦 Portable - Single database file, no external dependencies

Installation

npm install @ai-devkit/memory

Quick Start

As MCP Server

Add to your MCP client configuration:

{
  "mcpServers": {
    "memory": {
      "command": "npx",
      "args": ["@ai-devkit/memory"]
    }
  }
}

Using the Tools

Store Knowledge

{
  "tool": "memory.storeKnowledge",
  "arguments": {
    "title": "Always use Response DTOs for API endpoints",
    "content": "When building REST APIs, always use Response DTOs instead of returning domain entities directly. This provides better API versioning, security, and decoupling.",
    "tags": ["api", "backend", "dto"],
    "scope": "global"
  }
}

Search Knowledge

{
  "tool": "memory.searchKnowledge",
  "arguments": {
    "query": "building an API endpoint",
    "contextTags": ["api"],
    "limit": 5
  }
}

API Reference

memory.storeKnowledge

Store a new knowledge item.

| Parameter | Type | Required | Description | |-----------|------|----------|-------------| | title | string | ✅ | Short description (10-100 chars) | | content | string | ✅ | Detailed explanation in markdown (50-5000 chars) | | tags | string[] | ❌ | Domain keywords (max 10) | | scope | string | ❌ | global, project:<name>, or repo:<name> |

memory.searchKnowledge

Search for relevant knowledge.

| Parameter | Type | Required | Description | |-----------|------|----------|-------------| | query | string | ✅ | Natural language task description (3-500 chars) | | contextTags | string[] | ❌ | Tags to boost matching results | | scope | string | ❌ | Scope filter (project results prioritized) | | limit | number | ❌ | Max results (1-20, default: 5) |

Ranking Algorithm

Results are ranked using:

final_score = bm25_score × tag_boost + scope_boost

Where:
  bm25_score  = FTS5 bm25() with column weights (title=10, content=5, tags=1)
  tag_boost   = 1 + (matching_tags × 0.1)
  scope_boost = +0.5 if scope matches, +0.2 if global

Database Location

Default: ~/.ai-devkit/memory.db

Development

# Install dependencies
npm install

# Build
npm run build

# Run tests
npm test

# Run with MCP Inspector
npm run inspect

# Start server
npm run start

License

MIT