Claude Mneme

Working with AI coding assistants often means repeatedly explaining context at the start of each session. Claude Mneme solves this by giving Claude Code persistent memory across sessions.

Named after the Greek goddess of memory, Mneme automatically captures prompts, task progress, git commits, and responses through Claude Code's hook infrastructure. When you start a new session, all that context is already there – no manual reloading required.

How It Works

The plugin operates through Claude Code's built-in hooks:

Everything gets stored in project-specific memory files under ~/.claude-mneme/projects/. Each project maintains its own isolated context, so switching between codebases doesn't create confusion.

Smart Summarization

Memory logs grow quickly. When your log reaches 50 entries, Mneme automatically summarizes the older entries using Claude Haiku, keeping the 10 most recent for quick access. This compression preserves important context while keeping the memory footprint manageable.

Key Features

• Zero setup – Works immediately with sensible defaults
• Project isolation – Separate memory stores for each codebase
• Manual memory – /remember command for explicit preference and fact storage
• Noise filtering – Automatically skips short prompts, confirmations, and duplicates
• Non-blocking - Async logging hooks don't slow down your workflow
• Configurable – Adjust summarization thresholds and model selection
• Token efficient – Designed to minimize costs: no MCP overhead, extractive summarization for agent output (no AI needed), cheaper Haiku model for AI summarization, user-configurable summarization intervals, and compact memory injection into context
• No API key required – Automatically uses your existing Claude subscription for summarization

Use Cases

Claude Mneme is particularly useful for:

• Long-running projects – Maintain continuity across weeks or months of development
• Coding preferences – Remember your style choices, naming conventions, and architectural decisions
• Specialized agents – Track work done by subagents across sessions
• Institutional knowledge – Build up understanding of complex project requirements over time

The result is an AI assistant that actually remembers your project - its history, your preferences, and the decisions you've made together.

Related Projects

• Vexscan – Security scanner that can vet Claude Code plugins like this one before installation
• Claude Simple Status – Companion statusline plugin showing model, context, and quota usage
• TokenLean – CLI toolkit for token-efficient code analysis, complementing Mneme's token-conscious design