How does User Memory extract my preferences?

It uses pattern matching via hooks that run after Claude responses to identify phrases regarding your tech stack, tools, and preferred coding style, then saves them to a local JSON profile.

Does it store my data in the cloud?

No, all memory files—including your profile, changelog, and session data—are stored locally on your machine in the ~/.claude/user-memory/ directory.

Can I manually manage or delete my stored data?

Yes, the data is stored in standard JSON format. You can manually edit the files, or use MCP tools like clear_user_profile and remove_preference for granular control.

What is the difference between Minimal and MCP modes?

Minimal mode uses lightweight shell scripts and jq for basic extraction, while MCP mode uses a Node.js server to provide real-time tools, confidence tracking, and advanced session continuity.

How does the preference decay system work?

In MCP mode, preferences have a confidence score that decreases over time if not reinforced. If a preference isn't mentioned for an extended period, it is automatically pruned to keep your profile accurate.

User Memory for Claude Code

Name: User Memory for Claude Code
Author: cameronsjo

bycameronsjo

•

Productivity & Workflow

Persists user preferences, tech stacks, and session context across Claude Code interactions to provide a personalized, long-term development profile.

The User Memory skill transforms Claude Code into a context-aware assistant by building a persistent profile of your development habits and technical preferences. It automatically extracts information such as your preferred tech stack, editor, and communication tone from your conversations without requiring manual input. By utilizing hooks and an optional MCP server, it manages session continuity, tracks project decisions, and implements a confidence-based decay system to ensure your profile remains relevant. This eliminates the need to repeat setup instructions and allows Claude to maintain a consistent 'memory' of how you work across different sessions and projects.

Key Features

01Automatic preference extraction from conversation patterns with zero latency.

02Dual-mode implementation offering lightweight shell scripts or a full TypeScript MCP server.

03Session continuity tracking for task progress, technical decisions, and cross-session summaries.

04Confidence-based decay system that automatically phases out stale or outdated preferences.

05Detailed changelog and audit trail for all profile modifications and extractions.

061 GitHub stars

Use Cases

01Maintaining consistent coding standards and editor preferences across multiple disjointed sessions.

02Resuming complex development tasks where specific architectural decisions need to be remembered.

03Personalizing AI response tone and detail level based on your specific professional role.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add cameronsjo/claude-marketplace user-memory

For use in Claude.ai and ChatGPT

Download Skill