MCP-Memos

MCP-Memos is an memo tool based on MCP that allows users to record and retrieve text information.

It's perfect for developers to quickly save and find information in their workflow(may be cursor, etc) without switching to other applications.

🔍 Advanced LLM-Powered Search

MCP-Memos uses large language models for retrieval, providing the most powerful fuzzy search capability available:

Semantic understanding: Find content based on meaning, not just keywords
Context-aware: Understands what you're looking for even with incomplete descriptions
Natural language queries: Search as you would ask a human, no special syntax needed
Conceptual matching: Retrieves information by understanding concepts, not just text matching

Unlike traditional vector or text-based fuzzy search, MCP-Memos leverages the full power of LLMs to truly understand your retrieval intent, making it the most effective information retrieval approach available today.

Tool Documentation

| Tool Name | Description | Input Parameters | |-----------|-------------|------------------| |store_memo| Save important text information and add tags for easy retrieval later| tag:keyword or tag or description <br> content: that you want to save | |retrieve_memo| Retrieve previously saved text content based on keywords | text keyword or tag or description |

How to Use

Installation

Download the mcp-memos binary file according to your computer's architecture from releases

Configuration

Add MCP-Memos to the macPilotCli configuration file

{
    "mcpServers": {
      "MCP-Memos":{
        "command": "path/to/mcp-memos",
        "env": {
          "LLM_TOKEN": "xxxxxx",
          "LLM_BASE_URL": "xxxxxxxx/v1",
          "ANTHROPIC_MODEL": "xxxxx"
        }
      }
    }
}

Basic Usage

Record information

When you need to save information during development without switching to another app:

record this memo. its {description}

{content}

Retrieve information

When you need to find previously saved information:

Find this memo about {description}

[!NOTE] Descriptions can vary as MCP-Memo's LLM understands concepts, not just keywords, matching similar content regardless of wording.

LLM_BASE_URL optional, default is https://api.anthropic.com

ANTHROPIC_MODEL optional, default is claude-3-7-sonnet-20250219

Will accelerate the implementation of the sampling feature, which will eliminate the need to configure all environment parameters