← 返回
未分类

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...
Install and configure the LLM-WikiMind MCP server — a local knowledge base built on Karpathy's LLM Wiki pattern. Triggers: install wikimind, setup knowledge...
hal-9909 hal-9909 来源
未分类 clawhub v1.0.0 1 版本 100000 Key: 无需
★ 0
Stars
📥 299
下载
💾 0
安装
1
版本
#bm25#karpathy#knowledge-base#latest#llm-wiki#local-first#mcp#second-brain#setup#wiki

概述

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

pip3 install qmd

Step 2: Clone the repo

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:

./wikimind init

Options:

CommandWhen to use
----------------------
./wikimind initCreate a fresh wiki (interactive, asks for path)
./wikimind init ~/my-notes --adoptAdopt 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:

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

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

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

./wikimind start

Auto-start on login:

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

Step 6: Verify

./wikimind status

Expected output:

→ 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:

ToolDescription
-------------------
wiki_searchBM25 full-text search across all domains
wiki_getRead a specific page in full
wiki_listList pages by domain and/or type
wiki_ingest_noteWrite a new page + rebuild index + sync cache
wiki_domainsList 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:

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

版本历史

共 1 个版本

  • v1.0.0 当前
    2026-05-07 21:05 安全 安全

安全检测

腾讯云安全 (Keen)

安全,无风险
查看报告

腾讯云安全 (Sanbu)

安全,无风险
查看报告

🔗 相关推荐

knowledge-management

Wikimind Skill

hal-9909
Ingest articles, docs, notes, and web pages into your local LLM-WikiMind knowledge base (Karpathy's LLM Wiki pattern). T
★ 0 📥 465
ai-agent

self-improving agent

pskoett
记录自身发现以实现自我改进的技能
★ 4,156 📥 929,585
ai-agent

Agent Browser

rez0
用于 AI 代理的浏览器自动化 CLI。当用户需要与网站交互(包括浏览页面、填写表单、点击按钮、截图等)时使用。
★ 862 📥 341,615