Introduction
KnowFlow turns the articles and notes you save into a linked knowledge base. Your coding agent does the synthesizing; every page is a plain Markdown file you own.
If you have a bookmarks folder full of “read later” links, KnowFlow is the tool that finally makes them useful.
What it is
Section titled “What it is”- A place for everything you save. Articles, tweets, notes — captured as-is, unchanged, in a
raw/folder. - A wiki that builds itself — with your supervision. Any coding agent (ZCode, Codex CLI, Claude Code, OpenCode…) turns raw saves into linked pages — entities, concepts, comparisons — by following one written synthesis contract. Everything lands as
pendinguntil you approve it. - A map of what you know. The links between pages weave an interactive graph you can explore.
What it is not
Section titled “What it is not”- Not a one-command “AI writes your knowledge base” machine. The agent drafts; you review — one word per page. That boundary is the product.
- Not a note-taking app. There is no editor UI (yet). Your editor, file manager, or Obsidian works fine — everything is plain Markdown.
- Not a cloud service. It runs on your machine. Nothing leaves it unless you opt into semantic search.
Who it is for
Section titled “Who it is for”- People who save lots of links and feel the pile growing faster than their memory
- PKM folks (Obsidian/Notion users) who want their base to build itself instead of demanding manual gardening
- Anyone already living with an AI assistant who’d rather delegate the organizing than the judgment
How to read these docs
Section titled “How to read these docs”| Page | For |
|---|---|
| Quickstart | Everyone — a three-command walkthrough |
| Use with an AI assistant | Everyone — the core workflow, no scripting |
| Commands | Reference for every command |
| Architecture & Data model | The technically curious |
| Methodology | The “why” behind the design |
| FAQ | Quick answers |
Current limitations
Section titled “Current limitations”- URL capture relies on Jina Reader; YouTube and some logged-in platforms may need extra tooling.
- Semantic search needs a prebuilt vector index and an embedding API key.
- KnowFlow ships no LLM: synthesis runs in your external coding agent, following the written contract — KnowFlow never runs the agent for you.
See the roadmap in the README for what’s planned.