linehaul-ai-libsql-memory

libsql-memory Plugin for Claude Code

Persistent semantic memory for Claude Code sessions using LibSQL with vector search.

Quick Start

# Clone the repo
git clone https://github.com/linehaul-ai/libsql-memory.git
cd libsql-memory

# Run Claude Code with the plugin
claude --plugin-dir .

That's it! The plugin will automatically:

• Start the MCP server with memory tools
• Load automation hooks for context retrieval
• Enable /memory-status and /memory-clear commands

Prerequisites

Nomic embedding server (text-embedding-qwen3-embedding-8b) running at http://192.168.128.10:1234 (default)

Or use local embeddings (no external dependencies):

export LIBSQL_MEMORY_EMBEDDING_PROVIDER=local
claude --plugin-dir /path/to/libsql-memory

Or use OpenAI:

export LIBSQL_MEMORY_EMBEDDING_PROVIDER=openai
export OPENAI_API_KEY=your-key
claude --plugin-dir /path/to/libsql-memory

Alternative Installation

Copy to your project as a local plugin:

# Copy entire plugin directory
cp -r /path/to/libsql-memory /your/project/libsql-memory

# Then run Claude with the plugin directory
cd /your/project && claude --plugin-dir ./libsql-memory

Or install globally:

claude plugins add https://github.com/linehaul-ai/libsql-memory.git

Components

MCP Server

Provides memory tools:

• memory_store - Store a memory with content and namespace
• memory_retrieve - Get a specific memory by ID
• memory_search - Semantic search across memories
• memory_list - List memories in a namespace
• memory_delete - Delete a memory
• memory_stats - Get storage statistics

Hooks

Hook	Purpose
SessionStart	Loads project context and preferences at session start
UserPromptSubmit	Searches memory for context relevant to user queries
PreToolUse (Write/Edit)	Retrieves file-specific patterns before code changes
PostToolUse (Write/Edit)	Stores architectural decisions and patterns after changes
Stop	Saves session summaries and important learnings

Commands

• /memory-status - View memory statistics and recent entries
• /memory-clear [namespace] - Clear memories from a namespace

Skills

• memory-usage - Best practices for when and how to use memory effectively

Configuration

Environment Variables

Variable	Default	Description
LIBSQL_MEMORY_EMBEDDING_PROVIDER	nomic	Embedding provider: nomic, openai, local
LIBSQL_MEMORY_EMBEDDING_ENDPOINT	http://192.168.128.10:1234/v1/embeddings	Nomic server endpoint
OPENAI_API_KEY	-	Required if using OpenAI provider

Database

Stored at .claude/memory.db relative to where Claude Code runs.

Supported Platforms

Pre-built binaries are included for:

• macOS (Apple Silicon / arm64)

For other platforms, build from source:

cd src/plugin
make build

Building from Source

cd src/plugin
make build          # Build for current platform
make test           # Run tests
make verify         # Run all checks

Troubleshooting

1. MCP server not starting: Check binary exists in bin/ directory
2. Embedding errors: Ensure Nomic server is running or switch to local provider
3. Database errors: Ensure .claude/ directory exists and is writable
4. Platform not supported: Build from source using make build

License

MIT