# projectmem MCP server

> Provides a local-first memory and judgment layer for AI coding agents to prevent repeating failed fixes and retain project context across sessions.

- Repository: https://github.com/riponcm/projectmem
- Page: https://hostedmcp.org/servers/riponcm/projectmem
- Transport: stdio
- Stars: 881
- License: MIT
- Latest version: 0.3.5

## Install

### Claude Code

```sh
claude mcp add projectmem -- uvx projectmem
```

### Cursor (`~/.cursor/mcp.json`)

```json
{
  "mcpServers": {
    "projectmem": {
      "command": "uvx",
      "args": [
        "projectmem"
      ]
    }
  }
}
```

### VS Code (`.vscode/mcp.json`)

```json
{
  "servers": {
    "projectmem": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "projectmem"
      ]
    }
  }
}
```

### Claude Desktop (`claude_desktop_config.json`)

```json
{
  "mcpServers": {
    "projectmem": {
      "command": "uvx",
      "args": [
        "projectmem"
      ]
    }
  }
}
```

### Codex (`~/.codex/config.toml`)

```toml
[mcp_servers.projectmem]
command = "uvx"
args = ["projectmem"]
```

### Gemini CLI (`~/.gemini/settings.json`)

```json
{
  "mcpServers": {
    "projectmem": {
      "command": "uvx",
      "args": [
        "projectmem"
      ]
    }
  }
}
```

## FAQ

### What is coding agent memory?

It is a persistent record of issues hit, approaches attempted, fixes that worked, and decisions made, allowing AI agents to read this context at the start of a new session.

### How do I prevent my AI from repeating failed debugging attempts?

Projectmem uses typed events and a pre-commit warning system that fires before you repeat an approach that has already failed in the project.

### Which AI clients are supported?

The MCP server exposes tools to Claude Code, Claude Desktop, Cursor, Antigravity, and Codex.