seojoonkim

seojoonkim / memkraft

Public

Ultimate zero-dependency compound knowledge system for AI agents. Auto-extract, classify, search, and maintain memory in plain Markdown.

42
11
100% credibility
Found Apr 13, 2026 at 42 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

MemKraft is a lightweight tool for building and maintaining a personal knowledge base in plain text files, designed for easy extraction, searching, and organization of facts for AI agents or personal use.

How It Works

1
🧠 Discover MemKraft

You hear about a friendly tool that acts like a personal brain, helping you capture, organize, and recall important facts from your life without any hassle.

2
📁 Create your memory space

You make a simple folder on your computer where all your thoughts, notes, and facts will live safely in everyday text files you fully own.

3
✨ Capture facts easily

You share a news snippet or chat, and it automatically spots key people, companies, dates, and ideas, filing them neatly with sources and confidence tags.

4
🔍 Find info instantly

You ask about a person or topic, and it pulls up the best matches with smart fuzzy searching and context-aware rankings, like having a helpful assistant.

5
📋 Prep summaries

Before a meeting, you get a quick roundup of everything known about someone, including timelines and connections, ready to use right away.

6
🛠️ Tidy and heal

You run a gentle cleanup that checks for issues, removes duplicates, resolves mix-ups, and keeps your memory fresh and reliable over time.

🚀 Knowledge compounds

Your personal brain grows smarter with every use, connecting ideas, avoiding past mistakes, and making you more insightful every day.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 42 to 42 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is memkraft?

Memkraft is a Python CLI tool and library that builds a zero-dependency memory system for AI agents, storing everything in plain Markdown files you control. Pipe in text from news, chats, or logs; it auto-extracts entities, facts, and relationships across English, Korean, Chinese, and Japanese, then classifies and maintains them with confidence levels and source tags. Search fuzzy or agentic-style with context-aware ranking, generate meeting briefs, and run nightly "dream cycles" for deduping, decay, and health checks—no databases or APIs needed.

Why is it gaining traction?

Unlike Mem0 or Letta with their vector DBs and runtime deps, memkraft runs offline, git-friendly, on pure stdlib Python, making it dead simple to drop into any agent workflow. The killer hooks are feedback loops that compound knowledge (searches auto-file back to timelines), goal-weighted re-ranking (same query ranks differently by context), and "debugging is memory"—tracking hypotheses, evidence, and rejections to avoid repeat mistakes. Multilingual NER without LLMs seals it for global devs.

Who should use this?

AI agent builders integrating memory into langchain or custom bots, especially those handling CJK text or needing portable storage. Solo devs debugging code who want persistent hypothesis logs beyond scrollback. Researchers or VCs prepping entity briefs from scattered notes, like "Simon Kim VC Seoul."

Verdict

Grab it via pipx if you need lightweight agent memory—solid docs, 277 tests passing, MIT license, but at 42 stars and 1.0% credibility, it's alpha; test in non-prod first and watch for stability.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.