home / mcp / mcp standards mcp server

MCP Standards MCP Server

Self-learning AI standards system - learns from corrections and auto-updates CLAUDE.md

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "airmcp-com-mcp-standards": {
      "command": "uv",
      "args": [
        "run",
        "--directory",
        "/ABSOLUTE/PATH/TO/mcp-standards",
        "python",
        "-m",
        "mcp_standards.server_simple"
      ]
    }
  }
}

This MCP server learns your Claude preferences automatically and stores them locally so Claude applies remembered commands across sessions. It requires no manual correction steps after initial use, enabling you to set a preferred workflow and have Claude reuse it automatically in future interactions.

How to use

Use this local MCP server from your Claude client to remember your preferred commands and workflows. After you instruct Claude with a preference (for example, which package manager to use), Claude will automatically apply that preference in subsequent sessions without further prompts.

To take advantage of the automatic memory, run Claude with the MCP server configured as a local stdio process. The server learns from your corrections and applies them in later sessions.

Example usage patterns you can achieve after setup: - You tell Claude to use a specific tool or command, and Claude remembers it for future tasks. - Claude automatically selects your remembered workflow (for example, using a preferred package manager across Python projects). - Your remembered preferences persist across reboots and sessions without reconfiguration.

How to install

Prerequisites: you need Node.js and npm installed on your system. You will also run a Python module for the MCP server.

Follow these steps to install and set up the MCP server locally.

# 1) Clone the project repository
git clone https://github.com/airmcp-com/mcp-standards.git
cd mcp-standards

# 2) Install dependencies
npm install

# 3) Set up AgentDB (local vector memory)
npm run setup

Configuration snippet for Claude Desktop

Configure Claude Desktop to run the MCP server as a local process. Add the following to your Claude desktop configuration file. Replace the absolute path with the correct location on your system.

{
  "mcpServers": {
    "mcp-standards": {
      "command": "uv",
      "args": [
        "run",
        "--directory",
        "/ABSOLUTE/PATH/TO/mcp-standards",
        "python",
        "-m",
        "mcp_standards.server_simple"
      ]
    }
  }
}

Testing the memory

After starting Claude with the MCP server running, you can test memory via simple prompts that exercise recall. For example, declare a preference in one session and verify that Claude applies it automatically in the next session.

Notes and tips

- The memory is stored locally under your user profile, so you have full control over your data.

- If Claude does not apply a remembered preference, ensure the MCP server process is running and that the path to the server module is correct in your configuration.

Troubleshooting

If you encounter issues during setup, verify that Node.js is installed and at a compatible version, and check that the AgentDB setup completed successfully.

Look at logs for clues if Claude fails to connect to the MCP server. Confirm that the server is running and reachable via the configured command and path.

Security and privacy

Everything runs locally on your machine with no cloud calls. Your remembered preferences stay on your device and are used to customize Claude behavior in future sessions.

Available tools

remember

Store a user-remembered preference with a content string and a category

recall

Search and retrieve remembered preferences by query with top_k results

list_categories

List all detected memory categories such as python, javascript, git, docker, testing, general

memory_stats

Get statistics about stored memory and usage