LLM Wiki plugin
A plugin that turns PDFs, articles and notes into a shared markdown wiki that agents search, cite and keep growing
- 123 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Writes pages and environment scripts and downloads packages and a local embedding model at initialization.
- The shared Cloudflare wiki opens writes over MCP and a JSON API to agents with access.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/praneybehl-llm-wiki-pluginDetects 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.
- Writes pages and environment scripts and downloads packages and a local embedding model at initialization.
- The shared Cloudflare wiki opens writes over MCP and a JSON API to agents with access.
This is third-party code. Review the repository files before installing.
What it does
LLM Wiki is a second memory for AI agents. You add a source once, the agent turns it into linked markdown pages, then later finds the right section and answers with citations. Useful answers can be saved back into the wiki, so knowledge grows instead of being rebuilt. The wiki can be personal across all projects or live inside one project. In Claude Code the plugin installs a skill and ten `/wiki:*` commands: `init`, `ingest`, `query`, `lint`, `graph`, `learn`, `evolve`, `migrate`, `stats` and `upgrade`. Search is semantic and local by default, with no hosted vector database or embedding service. Version 4 added an optional shared wiki in your own Cloudflare account that agents on different machines reach over MCP or a JSON API.
Who it is for. Researchers, analysts and teams who accumulate articles, meeting notes and project decisions and want an agent to find them with citations.
Good fit when
- Articles and PDFs must become a knowledge base an agent cites in answers.
- Project decisions should live in readable markdown and not be retold to every new session.
- One wiki must serve Claude Code, Codex and Cursor.
- A team's knowledge across several machines should land in a shared wiki with atomic commits.
Not a fit when
- Your main problem is structured records and transactions: a regular database fits better.
- `uv` is missing: it is needed for script environments on every platform.
- Downloading a local embedding model at initialization is not allowed.
Example request
/wiki:ingest paper.pdf, then /wiki:query: which limitations of the method does the paper describe, with links to wiki pages.Limitations
The only prerequisite is `uv`. Wiki initialization installs mandatory components: FastEmbed, sqlite-vec, PyYAML and the local BAAI/bge-small-en-v1.5 model. The shared Cloudflare wiki is optional and deploys into your own account. Decision models for search gating and graph checks are off by default. The author's study of `/wiki:evolve` rejected both proposals and showed no quality gain, so treat that part as experimental. The latest commit is from October 2026.
How to disable. Run `/plugin uninstall llm-wiki@llm-wiki` and delete the wiki directory. If you deployed the shared wiki, delete your Cloudflare Worker separately.
FAQ
How do I install it in Claude Code?
With `/plugin marketplace add praneybehl/llm-wiki-plugin` and `/plugin install llm-wiki@llm-wiki`. The plugin works in any project and you choose the wiki location at initialization.
How do I install it in other agents?
Through the `skills` installer: `npx skills add praneybehl/llm-wiki-plugin -a <agent> -g`. Codex, Cursor, Gemini CLI, OpenCode, OpenClaw, Pi and others are supported.
Do I need an external vector database?
No. Semantic search is local by default on FastEmbed and sqlite-vec. Canonical content stays in plain markdown.
What does `/wiki:learn` do?
Captures verified experience from completed work. `/wiki:evolve` then proposes, tests, applies and rolls back skill changes based on it.
How do I move a local wiki to the shared one?
With `/wiki:migrate`: it moves the wiki, verifies every uploaded file by SHA-256 and lists skipped files in the report.
Discussion
No questions or reviews yet.
Related
PluginsLLM Wiki Manager keeps a personal Markdown wiki through an agent: it ingests sources, links pages and answers with citations
A Claude Code plugin for Obsidian: stores sources, compiles linked wiki pages and answers from the accumulated knowledge.
A workflow for Claude Code, Codex and Gemini CLI where the agent files source documents into a persistent, linked wiki
Claude Code plugin for a personal wiki per Karpathy's pattern: the agent grows pages from sources and answers with citations.
Turns session history of Claude Code, Codex, Cursor and Gemini CLI into a searchable wiki on Karpathy's LLM Wiki pattern, locally
A personal wiki in a .wiki folder: the agent reads sources, writes cross-linked pages and health-checks the knowledge base
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| LLM Wiki (this card) | no | 123 | ||
| LLM Wiki Memory | no | not required | 130 | agent memory shaped as a wiki with a different design; LLM Wiki keeps canonical content in markdown and offers a shared wiki on Cloudflare |
| LLM Wiki | no | 107 | a wiki following the LLM Wiki idea; this plugin adds ten commands, a graph and a shared version |
README in short
The README covers installation, why the plugin exists, version 4 features and sections on the shared wiki and the optional decision layer. Documentation is hosted on GitHub Pages.