Open source · Runs locally · Your data stays yours

Stop saving. Start knowing.

KnowFlow turns the articles and notes you save into linked knowledge pages — your AI assistant does the organizing, and every page is plain Markdown you own.

npx @jerryjiao/knowflow@latest init my-wikiOne-line install — npm i -g @jerryjiao/knowflow (or let your AI assistant run it)

How many “read later” links are sitting in your bookmarks right now?

Saving isn’t learning. Bookmarks solve storing — never remembering, never connecting.

Real screenshot

Your saves, grown into a map of what you know

Every page is a Markdown file; the links between them weave this graph automatically — zoom, drag, follow connections.

KnowFlow knowledge graph screenshot

Build your first graph in three steps →

What is KnowFlow

Three promises

📁

Every page is yours

Your knowledge base is plain Markdown in folders. Pack it up anytime — no lock-in, ever.

🤝

An AI assistant does the organizing

Claude, Cursor, or any assistant you trust reads your raw saves, drafts linked pages, and waits for your review.

🕸️

See how knowledge connects

Linked pages weave themselves into a graph, surfacing connections you never searched for.

How it works

Save, organize, see

Three steps, fully transparent, no black box.

1

Save

Articles, tweets, and notes land in raw/ exactly as captured.

raw/
2

Organize

Your AI assistant drafts entity, concept, and comparison pages from templates — you review.

wiki/
3

See

One command builds the graph, checks links, and groups by tag.

graph.html
For the technically curious: the whole loop is three commands
$knowflow init my-wiki
$knowflow ingest “that article you saved months ago” --source text
$knowflow graph --no-open ✓ graph/graph.html

Peace of mind

The AI organizes. It never decides.

The assistant drafts every page, but it cannot touch your original saves — what you captured stays exactly as captured.

Every change is a visible Markdown edit you can revert, and swapping to a different assistant changes nothing.

Inspired by Andrej Karpathy’s LLM Wiki: knowledge curated into connected pages beats a pile of saved links.

Also included

Thoughtful details

📁

Plain files, no lock-in

Everything is local Markdown — you can open it in Obsidian directly.

🤖

Any AI assistant

If it can read and write files, it can run your wiki. Swap anytime.

🕸️

Knowledge graph

Interactive viewer plus raw graph data you can export anywhere.

🏥

Health checks

Broken links, empty pages, and orphans are caught — one command fixes them.

🏷️

Tag hubs

Tag pages with [[tags]] and topic index pages build themselves.

🔎

Semantic search (optional)

Only when you want “find things like this” — off by default.

Start with your next “read later”

Three steps to your first knowledge graph.

Get started →npx @jerryjiao/knowflow@latest init my-wiki