Your second brain — built on plain Markdown, supercharged by AI, and wired into your daily workflow.

Most knowledge tools make you choose: either you get structure (Notion, Confluence) or you get speed (scratchpads, raw files). Knowledge OS gives you both. Plain Markdown files you own forever, semantic search that finds meaning not just keywords, and a growing library of AI skills that turn your notes into an active thinking partner.

Why people try it:

• You're drowning in bookmarks, tabs, and browser history you never revisit
• You take notes but can never find them when you need them
• You want AI assistance that actually knows your context, not just the internet
• You switch between tools (Notion → Obsidian → whatever) and lose continuity every time

Knowledge OS is a VS Code workspace + Claude Code setup that turns a folder of Markdown files into a personal knowledge platform. Everything runs locally. You own all your data.

The core loop:

1. Capture anything — articles, YouTube videos, meetings, ideas, research
2. Organize automatically with AI — frontmatter, tags, wikilinks, folder placement
3. Query by meaning — ask "what do I know about X?" and get synthesized answers from your own notes
4. Act — brainstorm new ideas, generate slides, prep for interviews, plan your week

Skills are specialized AI workflows that run inside Claude Code / Github Copilot. You can use it on cli as a developer, or just use extension on VSCode for Claude Code / Github Copilot to interact with the skill. Type a command or your message and the AI handles the rest.

• VS Code
• Claude Code extension
• Node.js 18+ and Python 3.11+

git clone https://github.com/coderphonui/knowledge-os.git
cd knowledge-os
./setup.sh

setup.sh symlinks skills into Claude Code, installs Python dependencies, and wires up git hooks that keep the search index current.

Open the project in VS Code, open the Claude chat panel, and type:

onboard me

The onboarding skill interviews you about your work context, goals, and preferences — then scaffolds your personal data/ folder and creates your profile. From this point, every skill is personalized to you.

Knowledge OS uses Markdown heavily. For a Notion-like reading experience, install any-markdown.editor from the VS Code marketplace.

The /query skill uses a local vector index (ChromaDB + Ollama) to find notes by meaning — not just keywords. No server, no Docker, no cloud API. Everything runs in-process on your machine.

brew install ollama
ollama pull bge-m3        # recommended: handles both Vietnamese and English
ollama serve

English-only alternative (faster, smaller):

ollama pull nomic-embed-text
export KB_EMBED_MODEL=nomic-embed-text

brain/skills/query/.venv/bin/python brain/skills/query/scripts/index.py full

After first setup the index stays current automatically — a Claude Code hook re-indexes after every write, and a file watcher catches edits from other editors.

source ./setup.sh env            # load into current session
source ./setup.sh env --install  # persist to ~/.zshrc

Tí Chat (brain/apps/tichat) is an optional Slack bot that lets you query your Knowledge OS from anywhere — your phone, a team channel, wherever Slack is open.

Send what do I know about LangChain? or save this: <url> and the agent synthesizes an answer from your own notes, projects, and journals.

• Runs in Socket Mode — no public URL or port forwarding needed
• Powered by Gemini API via LangGraph ReAct
• Loads every skill from brain/skills/ automatically at startup

See brain/apps/tichat/README.md for setup (Slack app creation, API keys, environment config).

Best for mobile access. Note: Tí Chat does not have a built-in web search tool — for research-heavy sessions, Claude Code in VS Code is the better choice.

data/                        # Your personal knowledge — all plain Markdown
├── _templates/              # topic, project, reference, brainstorm, journal
├── topics/                  # Evergreen concept notes
├── projects/{slug}/         # overview.md, decisions.md, roadmap.md
├── journal/                 # Daily and weekly entries
├── references/              # Captured articles, videos, papers
└── profile.md               # Your profile — personalizes every skill

brain/skills/                # AI skill library
├── query/                   # Semantic search (ChromaDB + Ollama)
├── capture-reference/       # URL and article capture
├── youtube-notes/           # Video transcript + analysis
├── brainstorm/              # Thinking partner
├── weekly-review/           # Weekly synthesis
├── calendar/                # Calendar management
└── ...                      # 15+ skills total

brain/apps/
└── tichat/                  # Slack bot — query your KB from any device

Your data/ structure is scaffolded to match your role and preferences during onboarding. You can always reorganize it later — the skills adapt.