Callosium — Claude Code MCP server: what it does, how to install
Connects your AI assistants to a single memory stored in local Markdown files. Keeps your notes private and stops repetitive project explanations.
claude mcp add --env CALLOSIUM_MODEL_DIR=<CALLOSIUM_MODEL_DIR> --transport stdio callosium -- npx -y callosiumWhat Callosium does
Callosium turns local Markdown notes or an Obsidian vault into a shared knowledge base for coding assistants. When asked a question, connected clients query this local memory first and cite the exact source files. If the answer is missing from your notes, the tool avoids guesswork and states that fact directly.
The architecture runs locally through a dashboard at http://localhost:4319. You control folder access per agent. The engine also signs every file update with the identity of the AI that wrote it.
Node.js 20 or newer is required. On its initial start, the tool fetches a 120MB model for English and Arabic semantic search. If no network connection is available, keyword matching continues to function on its own.
What it helps with
- share project context across different AI editors and chat apps
- connect local Obsidian vaults directly to coding assistants
- restrict specific folders from being viewed by selected agents
- verify statements through exact citations from local notes
Example prompts for the agent
- Look up our database deployment steps in my local notes
- Save the agreed API endpoint design into the project notes folder
- Check the vault for broken links and orphan files
How to install Callosium
Platform CLI
claude mcp add --env CALLOSIUM_MODEL_DIR=<CALLOSIUM_MODEL_DIR> --transport stdio callosium -- npx -y callosiumReplace <…> with your own values
Platform CLI
codex mcp add callosium --env CALLOSIUM_MODEL_DIR=<CALLOSIUM_MODEL_DIR> -- npx -y callosiumReplace <…> with your own values
Config file
{
"mcpServers": {
"callosium": {
"command": "npx",
"args": [
"-y",
"callosium"
],
"env": {
"CALLOSIUM_MODEL_DIR": "<CALLOSIUM_MODEL_DIR>"
}
}
}
}Replace <…> with your own values
Config file
{
"servers": {
"callosium": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"callosium"
],
"env": {
"CALLOSIUM_MODEL_DIR": "<CALLOSIUM_MODEL_DIR>"
}
}
}
}Replace <…> with your own values
Platform CLI
gemini mcp add -e CALLOSIUM_MODEL_DIR=<CALLOSIUM_MODEL_DIR> callosium npx -y callosiumReplace <…> with your own values
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 Callosium
How to install Callosium in Claude Code?
Run the command claude mcp add --env CALLOSIUM_MODEL_DIR=<CALLOSIUM_MODEL_DIR> --transport stdio callosium -- npx -y callosium in your terminal, replacing <CALLOSIUM_MODEL_DIR> with your chosen model folder.
Does Callosium send notes to an external cloud?
All notes remain in standard Markdown files on your local drive. The free app requires no account, and your files stay readable even if you remove the tool.
What dependencies are required before running Callosium?
You need Node.js 20 or newer on your machine. The first launch also downloads an approximate 120MB model to power semantic search in English and Arabic.
