LLM Wiki — a Claude Code Plugin
Build and maintain an LLM-curated personal knowledge base in your project. An implementation of Andrej Karpathy's LLM Wiki pattern as a Claude Code plugin, designed to scale to thousands of pages without becoming a context bottleneck.
What is this?
Most ways of using LLMs with documents look like RAG: you upload files, the LLM retrieves chunks at query time, generates an answer, and nothing accumulates. Every question re-derives knowledge from raw fragments. Karpathy's LLM Wiki pattern flips this — when a new source arrives, the LLM compiles it once into a persistent, structured wiki of markdown pages, and subsequent queries read the pre-synthesized wiki rather than the raw sources. Knowledge compounds.
This plugin packages the pattern as a Claude Code skill plus six slash commands (/wiki:init, /wiki:ingest, /wiki:query, /wiki:lint, /wiki:stats, /wiki:graph) and a small set of bundled Python scripts (BM25 search, structural lint, stats with scaling thresholds, plus an optional compiled graph layer). You curate sources and ask questions; Claude does the bookkeeping.
Why use it
The maintenance burden is what kills personal wikis. Updating cross-references, keeping summaries current, noting when new data contradicts old claims, maintaining consistency across dozens of pages — humans abandon this work. LLMs don't get bored, don't forget to update a backlink, and can touch fifteen files in one pass. The wiki stays alive because the cost of maintenance is near zero.
The pattern shines for accumulating textual research over weeks or months — papers, articles, transcripts, meeting notes, book chapters, customer calls — and degrades for highly relational data where a real database would serve better.
Installation
Claude Code — full plugin
The native path: the skill, the five /wiki:* slash commands, and the marketplace manifest all ship in one install.
/plugin marketplace add praneybehl/llm-wiki-plugin
/plugin install llm-wiki@llm-wiki
Once installed, the plugin works in any project — the wiki itself lives in the project's working directory, not in the plugin.
Other coding agents — skill only
The llm-wiki skill uses the standard agentskills.io format, so it installs cleanly into any agent supported by the skills CLI. Pick the --agent flag that matches your setup:
# Install globally so the skill is available across all projects
npx skills add praneybehl/llm-wiki-plugin -a <agent> -g
# Or install into the current project only
npx skills add praneybehl/llm-wiki-plugin -a <agent>
| Agent | --agent value | Invoke via | Scripts run |
|---|
| Claude Code | claude-code | /wiki:* slash commands (bundled) or natural language | ✅ |
| Codex (OpenAI) | codex | /skills or $llm-wiki / natural language | ✅ |
| Cursor | cursor | /llm-wiki or natural language | ✅ |
| Gemini CLI | gemini-cli | /skills management commands / natural language | ⚠️ unverified |
| OpenCode | opencode | natural language (agent invokes the native skill tool) | ⚠️ unverified |
| OpenClaw | openclaw | auto-exposed as a user command | ⚠️ scripts don't auto-execute |
| Pi Agent | pi | /skill:llm-wiki or natural language | ✅ |
OpenCode also reads .claude/skills/ and ~/.claude/skills/, so if you already installed the skill for Claude Code you can use it in OpenCode without a second install.
Hermes Agent (Nous Research) and other agentskills.io-compatible runtimes that aren't yet in the npx skills registry can still use this skill — clone the repo and symlink or copy skills/llm-wiki/ into the agent's skills directory (e.g. ~/.hermes/skills/llm-wiki/).
git clone https://github.com/praneybehl/llm-wiki-plugin.git
ln -s "$(pwd)/llm-wiki-plugin/skills/llm-wiki" ~/.hermes/skills/llm-wiki
A few things to know when using the skill outside Claude Code:
- Slash commands are Claude Code-only. The five
/wiki:* commands live in commands/wiki/ as Claude Code plugin manifests. In other agents, invoke the skill by natural language ("add this paper to the wiki", "what does the wiki say about X", "lint the wiki") — the SKILL.md handles the rest.
- Script execution varies. OpenClaw doesn't run bundled Python scripts the way Claude Code does, so BM25 search, lint, and stats degrade to model-driven behavior. Gemini CLI and OpenCode don't document their script-execution semantics explicitly — test before relying on the bundled scripts in those environments. For the full feature set, use Claude Code, Codex, Cursor, or Pi.
- The wiki itself is agent-agnostic. It's just a directory of markdown files. You can ingest with one agent and query with another; nothing in
wiki/ ties it to a specific runtime.
Quick start
