LLM-WikiMind MCP Setup

Install and configure the LLM-WikiMind MCP server — a local knowledge base built on Karpathy's LLM Wiki pattern. Triggers: install wikimind, setup knowledge base MCP, configure wiki-kb, llm-wikimind setup, 安装知识库 MCP, 配置 wiki-kb, 搭建个人知识库.

LLM-WikiMind MCP Setup Skill

Set up the LLM-WikiMind MCP server — a local knowledge base built on Karpathy's LLM Wiki pattern.

What you get: 5 MCP tools (wiki_search, wiki_get, wiki_list, wiki_ingest_note, wiki_domains) backed by BM25 full-text search over your local Markdown files. No embeddings, no vector DB, no cloud.

Trigger phrases

• "Install WikiMind" / "Set up my knowledge base MCP"
• "Configure wiki-kb" / "llm-wikimind setup"
• "安装知识库 MCP" / "配置 wiki-kb" / "搭建个人知识库"

---

Installation Workflow

Step 1: Install the only dependency

bash

pip3 install qmd

Step 2: Clone the repo

bash

git clone https://github.com/HAL-9909/llm-wikimind
cd llm-wikimind

Step 3: Initialize the wiki

Run the interactive setup — it will ask where to store the knowledge base and handle everything else:

bash

./wikimind init

Options:

Command	When to use
./wikimind init	Create a fresh wiki (interactive, asks for path)
./wikimind init ~/my-notes --adopt	Adopt an existing Markdown directory

init automatically:

• Creates the standard wiki directory structure
• Copies the MCP server and watcher into the wiki
• Builds the initial BM25 search index
• Prints the exact config snippet to paste into the AI client

Step 4: Register the MCP server

The init command prints the exact snippet. For reference:

CatDesk / OpenClaw:

bash

catdesk mcp add --name wiki-kb --json '{
  "command": "python3",
  "args": ["<WIKIMIND_ROOT>/.wiki-mcp/server.py"],
  "env": {"WIKIMIND_ROOT": "<WIKIMIND_ROOT>"}
}'

Replace <WIKIMIND_ROOT> with the path chosen in Step 3 (e.g. ~/Documents/wiki).

Claude Desktop — add to ~/Library/Application Support/Claude/claude_desktop_config.json:

json

{
  "mcpServers": {
    "wiki-kb": {
      "command": "python3",
      "args": ["<WIKIMIND_ROOT>/.wiki-mcp/server.py"],
      "env": { "WIKIMIND_ROOT": "<WIKIMIND_ROOT>" }
    }
  }
}

Step 5: Start the auto-sync watcher

bash

./wikimind start

Auto-start on login:

bash

echo '/path/to/llm-wikimind/wikimind start > /dev/null 2>&1' >> ~/.zshrc

Step 6: Verify

bash

./wikimind status

Expected output:

text

→ Wiki root:  ~/Documents/wiki
✓ Wiki exists — 1 domain(s), 3 pages
✓ Watcher running (pid 12345)
✓ qmd installed

---

MCP Tools Reference

Once registered, these 5 tools are available to any MCP-compatible AI client:

Tool	Description
wiki_search	BM25 full-text search across all domains
wiki_get	Read a specific page in full
wiki_list	List pages by domain and/or type
wiki_ingest_note	Write a new page + rebuild index + sync cache
wiki_domains	List all registered domains and their trigger keywords

---

Auto-update: always in sync

The watcher runs in the background and keeps everything current automatically:

• Add a new domain folder → detected within 10 seconds → MCP tool descriptions updated
• Edit DOMAIN.md keywords → AI knows about the change in the next conversation
• AI writes a new page via wiki_ingest_note → index rebuilt automatically

No restarts. No manual config changes.

---

Adding knowledge after setup

Install the companion ingest skill:

bash

npx clawhub@latest install wikimind-ingest

Then just say: "Add this to my knowledge base: [paste content or URL]"

---

Troubleshooting

qmd: command not found → Run pip3 install qmd

MCP server not responding → Check ./wikimind status and ensure WIKIMIND_ROOT is set correctly

Watcher not running → Run ./wikimind start; add to ~/.zshrc for auto-start

New domain not detected → Ensure DOMAIN.md has a keywords frontmatter field; check ./wikimind status