llm-wiki

A headless wiki engine for agents. 22 MCP tools. One Rust binary. No LLM inside.

Build knowledge that compounds — not answers that evaporate.

A git-backed Markdown wiki — searchable, typed, graph-linked. Accessible from the command line, from any MCP-compatible agent, or from any IDE via ACP.

The problem with RAG

Most AI knowledge tools retrieve and generate on every query. Each answer is disposable — nothing is learned, nothing is kept. Ask the same question twice and the LLM reasons from scratch.

llm-wiki implements a different pattern — the Dynamic Knowledge Repository (DKR), introduced by Andrej Karpathy:

Process sources at ingest time, not query time. The LLM integrates each source into the wiki — updating concept pages, creating source summaries, flagging contradictions — and commits the result. Knowledge compounds with every addition.

How it works

The engine is pure infrastructure. It manages files, git, full-text search, and graph structure. The LLM is always external — it calls the engine's tools via MCP, reads pages, writes pages, and commits knowledge. Intelligence flows through skills, not the binary.

LLM agent
  │
  ├── wiki_list(format: "llms")             → all pages grouped by type
  ├── wiki_search("mixture of experts")     → ranked results + facets
  ├── wiki_content_read("concepts/moe")     → full page + backlinks
  ├── wiki_graph(root: "concepts/moe")      → typed graph in Mermaid/DOT
  ├── wiki_suggest("concepts/moe")          → pages worth linking
  ├── wiki_content_new("concepts/new-page") → scaffold + returns local path
  ├── [write directly to path]              → no MCP round-trip
  └── wiki_ingest(path: "concepts/")        → validate, index, commit

A wiki page is a plain Markdown file with typed frontmatter:

---
type: concept
title: Mixture of Experts
status: active
confidence: 0.9
tags: [routing, scaling, efficiency]
sources:
  - sources/switch-transformer-2021
  - sources/mixtral-2024
concepts:
  - concepts/sparse-routing
  - concepts/scaling-laws
---

Sparse routing of tokens to expert subnetworks...

The engine validates frontmatter against a JSON Schema, extracts typed graph edges from sources and concepts, and indexes everything in tantivy. The graph is live the moment a page is committed.

Install

# macOS / Linux
curl -fsSL https://raw.githubusercontent.com/geronimo-iia/llm-wiki/main/install.sh | bash

# Windows (PowerShell)
irm https://raw.githubusercontent.com/geronimo-iia/llm-wiki/main/install.ps1 | iex

# Homebrew
brew install geronimo-iia/tap/llm-wiki

# Cargo
cargo install llm-wiki-engine

→ All installation options

Quick start

# Create a wiki space
llm-wiki spaces create ~/wikis/research --name research

# Start the MCP server
llm-wiki serve

Connect your agent or editor — VS Code, Cursor, Windsurf, Zed, Claude Code — via the MCP config. The 22 tools are immediately available.

→ Getting started guide · IDE integration

What agents can do

Full tool reference: docs/specifications/tools/

Skills

The engine exposes tools. Skills tell agents how to use them.

llm-wiki-skills is a Claude Code plugin that ships ready-to-use workflows:

Skills are plain Markdown files — readable by the LLM, replaceable, forkable. Write your own for your own workflows.

Architecture

llm-wiki-engine          pure Rust binary — tools, git, index, graph
llm-wiki-skills          Claude Code plugin — workflow skills (Markdown)
llm-wiki-hugo-cms        Hugo scaffold — render the wiki as a website

The engine has no opinions about workflows, LLM providers, or interfaces. Every LLM call happens outside the binary. Every workflow lives in a skill. The separation means skills ship independently, the engine stays stable, and nothing is coupled to a specific AI provider.

Technology

The file format is Markdown. The history store is git. Both predate llm-wiki and will outlive it — your wiki is readable, diffable, and portable with zero dependency on this tool. The engine itself is a single Rust binary with no runtime, no database, and nothing to keep running between sessions.

Single Rust binary. No runtime, no database, no Docker.

Documentation

Related Projects

Acknowledgments

• Andrej Karpathy — for the LLM Wiki gist that defined the Dynamic Knowledge Repository pattern.
• vanillaflava — for llm-wiki-claude-skills, which turned the pattern into a practical skill architecture.

llm-wiki is a continuation of agent-foundation.

Contributing

Contributing guide · Code of conduct · Security policy

License

MIT OR Apache-2.0