CROW: LLM Wiki — Claude Code MCP server: what it does, how to install
CROW: LLM Wiki is a Claude Code MCP server that acts as long-term memory backed by a local Markdown wiki.
AGPL-3.0 ★ 309 updated 2026-11-17
claude mcp add --env OKF_MCP_MODELS=<OKF_MCP_MODELS> --env OKF_BUNDLE=<OKF_BUNDLE> --transport stdio crow-llm-wiki -- uvx llmwiki2What CROW: LLM Wiki does
The server ingests documents, files them into folders, and answers questions with citations pointing directly to relevant notes. All data stays as plain Markdown files inside the ~/llm-wiki folder, so you can edit them directly or track changes in git.
Running llmwiki2 setup in the terminal configures the agent connection interactively. The system supports local inference via Ollama as well as API keys for OpenRouter and OpenAI.
Python version 3.11 or later is required, or the uv tool for isolated execution.
What it helps with
- organize documents into a wiki folder structure
- answer questions citing source notes
- connect persistent agent memory to Claude Code
- store all notes as plain Markdown files
Example prompts for the agent
- remember that the vendor payment term is 45 days
- add the offer document from downloads to the wiki
- move the pricing note to the suppliers folder
How to install CROW: LLM Wiki
Platform CLI
claude mcp add --env OKF_MCP_MODELS=<OKF_MCP_MODELS> --env OKF_BUNDLE=<OKF_BUNDLE> --transport stdio crow-llm-wiki -- uvx llmwiki2Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from llmwiki2 <command>
Platform CLI
codex mcp add crow-llm-wiki --env OKF_MCP_MODELS=<OKF_MCP_MODELS> --env OKF_BUNDLE=<OKF_BUNDLE> -- uvx llmwiki2Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from llmwiki2 <command>
Config file
{
"mcpServers": {
"crow-llm-wiki": {
"command": "uvx",
"args": [
"llmwiki2"
],
"env": {
"OKF_MCP_MODELS": "<OKF_MCP_MODELS>",
"OKF_BUNDLE": "<OKF_BUNDLE>"
}
}
}
}Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from llmwiki2 <command>
Config file
{
"servers": {
"crow-llm-wiki": {
"type": "stdio",
"command": "uvx",
"args": [
"llmwiki2"
],
"env": {
"OKF_MCP_MODELS": "<OKF_MCP_MODELS>",
"OKF_BUNDLE": "<OKF_BUNDLE>"
}
}
}
}Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from llmwiki2 <command>
Platform CLI
gemini mcp add -e OKF_MCP_MODELS=<OKF_MCP_MODELS> -e OKF_BUNDLE=<OKF_BUNDLE> crow-llm-wiki uvx llmwiki2Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from llmwiki2 <command>
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about CROW: LLM Wiki
How to install CROW: LLM Wiki in Claude Code?
Run claude mcp add llm-wiki-v2 -- uvx llmwiki2, followed by claude mcp add --env OKF_MCP_MODELS=<OKF_MCP_MODELS> --env OKF_BUNDLE=<OKF_BUNDLE> --transport stdio crow-llm-wiki -- uvx llmwiki2.
What does CROW: LLM Wiki do?
It turns ingested documents into a structured Markdown wiki and provides long-term memory for AI agents with cited references.
What is required to run CROW: LLM Wiki?
Python 3.11 or newer or the uv tool is required, alongside an AI provider such as Ollama or OpenRouter.