Claude Memory Extractor

Why it's ranked here

The pipeline is concrete and inspectable, with local transcript parsing, bounded chunking, and plain Markdown output. Human validation reported strong agreement after prompt calibration. Still, the repository’s own warning says the work received limited human scrutiny, and extracted lessons require manual review.

What's good

It preserves source context, distinguishes problem-solving methods from narrow technical fixes, and records uncertainty instead of presenting every inference as certain. Users can select a starting date and custom input or output directories. Incremental runs remember the previous extraction time, while readable Markdown keeps results easy to inspect and edit.

Tradeoffs

It requires Node.js 18 or newer plus an installed, authenticated Claude CLI. Each conversation reportedly takes one to two minutes and costs roughly $0.50 to $2.00 in tokens. There is no search interface or automatic memory injection. The advertised dry run still performs extraction and merely limits processing to twenty sessions.

How to use it well

It suits Claude Code users willing to periodically mine debugging sessions, corrections, and working preferences, then manually curate the resulting Markdown. Start with a narrow time range and inspect several outputs before processing a larger archive. It does not provide retrieval, semantic search, automatic context injection, or a complete memory-management system.

Observed

License

MIT

Primary language

TypeScript on Node.js

Install surface

npm package with global installation and source-build instructions

Interface

Command-line interface with extract and info commands

Runtime support

Node.js 18 or newer

External requirement

Installed and authenticated Claude CLI

Output format

Markdown files with YAML frontmatter

Runtime dependencies

Commander and dotenv