Wax
Single-file memory layer for AI agents, sub mili-second RAG on Apple Silicon. Metal Optimized On-Device. No Server. No API. One File. Pure Swift
What is Wax?
Wax is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to single-file memory layer for ai agents, sub mili-second rag on apple silicon. metal optimized on-device. no server. no api. one file. pure swift
Single-file memory layer for AI agents, sub mili-second RAG on Apple Silicon. Metal Optimized On-Device. No Server. No API. One File. Pure Swift
This server falls under the Knowledge & Memory category on MCPgee, the world's largest MCP server directory with 33,000+ servers.
Features
- Single-file memory layer for AI agents, sub mili-second RAG
Use Cases
Maintainer
Works with
Installation
Manual Installation
npx waxConfiguration
Configuration Details
claude_desktop_config.json
Performance
Response Metrics
Resource Usage
How to Set Up and Use Wax
Wax is a single-file, on-device memory layer for AI agents that delivers sub-millisecond hybrid (text + vector) search on Apple Silicon using Metal GPU acceleration and CoreML — no server, no API key, no network round-trip required. Memories are stored in a compact binary .wax file that can be synced across devices via iCloud or Dropbox, and the MCP server interface lets any MCP-compatible AI agent read and write persistent memory. Developers and power users on macOS choose it for fully private, offline-capable agent memory that keeps AI context fast without cloud dependencies.
Prerequisites
- macOS with Apple Silicon (M1/M2/M3/M4) — Intel Macs work via Rosetta but without Metal GPU acceleration
- Swift 6 or later if building the CLI from source
- Node.js and npx for the MCP server installation
- An MCP-compatible client such as Claude Desktop
- Sufficient disk space for the .wax memory file (typically a few MB)
Install the Wax MCP server
Use npx to install and run the waxmcp package as an MCP server at the user scope. This sets up Wax as a persistent MCP tool provider.
npx -y waxmcp@latest mcp install --scope userBuild the CLI from source (optional)
If you prefer the command-line tool for direct memory management outside of an MCP client, build and install the wax-cli binary using Swift.
git clone https://github.com/christopherkarani/Wax.git
cd Wax
swift build -c release
cp .build/release/wax-cli /usr/local/bin/Store a memory via CLI
Use wax-cli to save a fact or piece of information to the default memory store at ~/.wax/memory.wax.
wax-cli remember "The project uses PostgreSQL 16 with pgvector extension for embeddings."Search memories
Retrieve relevant memories using hybrid search (combines semantic vector similarity with lexical text matching) or text-only mode for exact keyword matching.
wax-cli search "database setup" --mode hybrid --topK 5
wax-cli search "PostgreSQL" --mode textRun as a daemon for always-on memory
Start the wax-cli daemon so AI agents can query memory in real time without cold-start latency. Specify a custom store path to control where the .wax file lives.
wax-cli daemon --store-path ~/.wax/memory.waxConfigure Claude Desktop to use Wax
Add the Wax MCP server to your Claude Desktop configuration so Claude can automatically read and write persistent memories during conversations.
Wax Examples
Client configuration
Configure Claude Desktop to connect to the Wax MCP server for on-device agent memory.
{
"mcpServers": {
"wax": {
"command": "npx",
"args": ["-y", "waxmcp@latest"],
"env": {
"WAX_STORE_PATH": "/Users/yourname/.wax/memory.wax"
}
}
}
}Prompts to try
Example prompts that exercise Wax's memory store and retrieval capabilities.
- "Remember that the API rate limit for our service is 1000 requests per minute"
- "What do you remember about our database architecture?"
- "Search your memory for anything related to authentication and summarize it"
- "Store the fact that we use Tailwind CSS v4 in this project"
- "Forget the note about the old Redis configuration"Troubleshooting Wax
Vector search returns no results on Intel Mac
Intel Macs lack the Apple Neural Engine and Metal shader support required for on-device vector embeddings. Use '--mode text' for lexical-only search, or migrate to an Apple Silicon Mac to enable hybrid and vector search modes.
wax-cli command not found after copying the binary
Ensure /usr/local/bin is in your PATH. Run 'echo $PATH' to verify, and add 'export PATH="/usr/local/bin:$PATH"' to your ~/.zshrc or ~/.bashrc if it's missing. Also confirm the binary is executable: 'chmod +x /usr/local/bin/wax-cli'.
Memory file grows very large over time
The .wax file is a single binary store — periodically prune old memories using 'wax-cli forget' for specific items, or delete ~/.wax/memory.wax to start fresh. Wax does not currently auto-compact, so manual pruning is needed for long-running agents.
Frequently Asked Questions about Wax
What is Wax?
Wax is a Model Context Protocol (MCP) server that single-file memory layer for ai agents, sub mili-second rag on apple silicon. metal optimized on-device. no server. no api. one file. pure swift It connects AI assistants to external tools and data sources through a standardized interface.
How do I install Wax?
Follow the installation instructions on the Wax GitHub repository. Clone the repo, install dependencies, and add the server config to your AI client.
Which AI clients work with Wax?
Wax works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.
Is Wax free to use?
Yes, Wax is open source and available under the Apache-2.0 license. You can use it freely in both personal and commercial projects.
Wax Alternatives — Similar Knowledge & Memory Servers
Looking for alternatives to Wax? Here are other popular knowledge & memory servers you can use with Claude, Cursor, and VS Code.
MemPalace
★ 52.6kA local AI memory system that stores all conversations verbatim and organizes them into navigable structures. It provides 19 MCP tools for AI assistants to search and retrieve past decisions, debugging sessions, and architecture debates automatically
Kratos
★ 25.7k🏛️ Memory System for AI Coding Tools - Never explain your codebase again. MCP server with perfect project isolation, 95.8% context accuracy, and the Four Pillars Framework.
Context Mode
★ 15.4kAn MCP server that preserves LLM context by intercepting large data outputs and returning only concise summaries or relevant sections. It enables efficient sandboxed code execution, file processing, and documentation indexing across multiple programm
Memu
★ 13.7kMemory for 24/7 proactive agents like OpenClaw.
MemOS
★ 9.3kMemOS (Memory Operating System) is a memory management operating system designed for AI applications. Its goal is: to enable your AI system to have long-term memory like a human, not only remembering what users have said but also actively invoking, u
Everos
★ 5.4kBuild, evaluate, and integrate long-term memory for self-evolving agents.
Browse More Knowledge & Memory MCP Servers
Explore all knowledge & memory servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.
Set Up Wax in Your Editor
Choose your AI client for step-by-step setup instructions.
Quick Config Preview
Add this to your claude_desktop_config.json or .cursor/mcp.json
Ready to use Wax?
Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.