llmwiki-cli
CLI for a personal Markdown knowledge base: the agent writes, searches and lints pages with wiki commands; the tool calls no model
- 104 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- `wiki write` and `wiki delete` change and delete the base files, so an agent mistake wipes pages
- The base folder sits on your disk in plain form and may contain personal notes
- The optional GitHub Pages graph workflow publishes the base content to the internet
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/llmwiki-cliDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- `wiki write` and `wiki delete` change and delete the base files, so an agent mistake wipes pages
- The base folder sits on your disk in plain form and may contain personal notes
- The optional GitHub Pages graph workflow publishes the base content to the internet
This is third-party code. Review the repository files before installing.
What it does
llmwiki-cli is a command-line program an agent uses to keep a knowledge base in the spirit of Andrej Karpathy's "LLM Wiki". It works on plain Markdown files on disk: reads, writes, deletes, searches and lints them. The model decides what to create and how to link pages, while the program itself calls no API. `wiki init` creates a folder with the config `.llmwiki.yaml`, an instruction file `SCHEMA.md` for the agent, a `raw/` folder for immutable sources and a `wiki/` folder for pages: entities, concepts, source summaries and cross-cutting analysis. `wiki write` takes JSON on standard input, turns it into a page with front matter and updates `wiki/index.md` itself. Then come `wiki search`, `wiki links`, `wiki backlinks`, `wiki orphans`, `wiki status` and `wiki lint`, which finds broken links, pages with no inbound links and pages without front matter. `wiki skill` prints a guide for the agent. You can keep several wikis and switch with `wiki use`.
Who it is for. Researchers and developers who accumulate notes and article summaries and want Claude Code or Codex to maintain a linked knowledge base in plain Markdown files.
Good fit when
- A folder of articles and notes has piled up and should become a linked base with summaries and cross-references
- The agent should answer questions from your materials and show which pages it relies on
- Once a week you want to find orphan pages and broken links before the base falls apart
- Different topics need separate bases that you can switch between easily
- You want to keep knowledge in Markdown under plain Git rather than in a closed note service
Not a fit when
- The base must sync with the cloud or Git on its own: there is no built-in sync
- You need vector search by meaning: `wiki search` searches the page text
- You work in Obsidian and want a plugin with a graphical interface: this is a console tool
- Page structure must not depend on agent discipline: the agent maintains it by following `SCHEMA.md`
Example request
Create a wiki in ~/wikis/ml about machine learning, write a summary of Attention Is All You Need and link it to a page about transformersLimitations
The program cannot collect sources by itself: the agent or you bring articles and documents into the `raw/` folder. There is no built-in sync with Git or a cloud, it does not run `git init`, and you keep history in ordinary Git. Installation needs Node.js 18 or Bun. Pages are written through a JSON command on standard input, so give the agent a direct instruction to read the `wiki skill` guide. The package version is 1.0.1.
How to disable. Remove the program with `npm uninstall -g llmwiki-cli`. The wiki folders and the registry in `~/.config/llmwiki/` stay on disk and can be deleted by hand.
FAQ
How do I install llmwiki-cli?
Run `npm install -g llmwiki-cli`. It installs two commands: the main `wiki` and a fallback `llmwiki` in case `wiki` is taken on your system. Then create a base: `wiki init my-wiki --name "My Notes" --domain "research"`.
Does the program call models or the internet?
No. It only reads and writes Markdown on disk. Your agent, for example Claude Code or Codex, decides which pages to create when it runs the `wiki` commands.
What exactly does wiki lint check?
It finds broken links between pages, pages with no inbound links and pages without front matter. For separate checks use `wiki orphans` and `wiki backlinks <path>`, and `wiki status` gives the overall picture.
How do I keep several knowledge bases?
Create them with `wiki init` in different folders, see the list in `wiki registry` and pick the active one with `wiki use <id>`. For a single command use the `--wiki` flag, and `wiki search <query> --all` searches all bases at once.
Can I show a graph of page links?
Yes, but by hand. Copy `.github/workflows/wiki-viz.yml` and the scripts `scripts/build-graph.js` and `scripts/build-site.js` from the repository to the base root, set up a Git repository there and enable GitHub Pages. The workflow parses `[[wikilinks]]` on every push.
Discussion
No questions or reviews yet.
Related
CLI toolsTurns session history of Claude Code, Codex, Cursor and Gemini CLI into a searchable wiki on Karpathy's LLM Wiki pattern, locally
A workflow for Claude Code, Codex and Gemini CLI where the agent files source documents into a persistent, linked wiki
LLM Wiki
llm-wiki
Knowledge bases that AI compiles: topics, import, llm-wiki lint checks, the /wiki:ingest and /wiki:archive commands. A plugin for Claude Code and Codex
A skill that lets an agent read the LLM Wiki desktop app's knowledge base over its local HTTP API: search, pages and graph
An agent saves web pages, PDFs and scans into a local vault, builds linked concepts and answers from them with source references.
Karpathy LLM Wiki
karpathy-llm-wiki
A skill for a personal knowledge base the model maintains: ingest sources, compile cited articles, lint integrity
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| llmwiki-cli (this card) | no | 104 | |
| llmwiki | no | 400 | builds a searchable wiki from agent session history rather than keeping notes through commands |
| LLM Wiki Skill | no | 153 | reads the LLM Wiki desktop app base over a local HTTP API instead of storing pages itself |
README in short
The README explains the split of labour: the program is the hands, the model is the head. It shows the folder layout, commands for managing, reading, writing and checking the base, typical agent workflows, an optional link graph on GitHub Pages and support for several bases with a rule for choosing the active one.