memorybank

Persistent cross-session memory for AI agents. Your AI agent forgets everything between sessions — memorybank fixes that.

Works with any MCP-compatible AI tool: Claude Code, Cursor, Windsurf, and more.

Quick Start

# Use directly with npx (zero install)
npx @ura-dev/memorybank --help

# Or install globally
npm install -g @ura-dev/memorybank

As MCP Server

Add to your Claude Code config (~/.claude/settings.json):

{
  "mcpServers": {
    "memorybank": {
      "command": "npx",
      "args": ["-y", "@ura-dev/memorybank", "memorybank-mcp"]
    }
  }
}

Or in Cursor/VS Code (.cursor/mcp.json):

{
  "mcpServers": {
    "memorybank": {
      "command": "npx",
      "args": ["-y", "@ura-dev/memorybank", "memorybank-mcp"]
    }
  }
}

MCP Tools

| Tool | Description | |------|-------------| | memory_store | Save a new memory (user, project, feedback, reference, custom) | | memory_recall | Retrieve the most relevant memories for a query | | memory_list | List all memories, filter by type or tag | | memory_get | Get a specific memory by ID | | memory_update | Update an existing memory | | memory_delete | Delete a memory | | memory_search | Full-text search across all memories | | memory_stats | Statistics about stored memories |

CLI Usage

# Store a memory
memorybank store user "team-lead" "User is the team lead for the backend services team"

# Recall relevant memories
memorybank recall "user role"

# Search
memorybank search "TypeScript"

# List all memories
memorybank list
memorybank list --type=feedback

# Get a specific memory
memorybank get abc123def456

# Delete
memorybank delete abc123def456

# Stats
memorybank stats

# Use namespaces for project isolation
memorybank store project "api-rewrite" "Migrating from REST to GraphQL" --namespace=myproject

Memory Types

• user — About the person: role, preferences, expertise
• project — About the work: goals, decisions, architecture
• feedback — Corrections and validated approaches
• reference — Pointers to external resources
• custom — Anything else

How It Works

• Zero dependencies — uses only Node.js built-in modules
• File-based storage — each memory is a markdown file with YAML frontmatter
• Auto-indexed — fast lookups via maintained index
• Namespaced — separate memory banks per project
• Secure — path traversal protection, input validation

Storage location: ~/.memorybank/ (override with MEMORYBANK_DIR env var)

Programmatic API

const mb = require('@ura-dev/memorybank');

// Store
const { id } = mb.store({ name: 'api key location', type: 'reference', content: 'API keys are in .env.local' });

// Recall
const memories = mb.recall('api keys');

// Search
const results = mb.search('TypeScript');

Why memorybank?

AI coding agents are powerful but stateless. Every new session starts from zero. memorybank gives your agent persistent memory:

• Remembers your preferences across sessions
• Tracks project decisions so the agent doesn't re-ask
• Stores corrections so mistakes aren't repeated
• Saves references to external resources

Think of it as a knowledge base that your AI agent builds and queries automatically.

License

MIT