Teach your AI to build with CleanSlice

An MCP server that gives AI coding agents instant access to CleanSlice architecture docs, conventions, and best practices.

Docs
$

For Claude Code. Install for Cursor, Windsurf & others

Server running
💡
Get better results from your AI agent

Installing the MCP server is step one. To make sure your agent always consults CleanSlice docs before writing code, add an enforcement rule to your editor:

Add an enforcement rule — a small config file that tells the agent to call get-started and search before every task. See examples per editor below.

How it works

Three steps to structured AI output

Connect once, and every prompt is architecture-aware.

1
Install the MCP server
Add CleanSlice MCP to Cursor, Claude Code, VS Code, Windsurf, or any MCP client.
2
Add enforcement rules
A small config file ensures the agent always consults CleanSlice docs before writing code.
3
Get CleanSlice architecture
Every prompt produces code that follows CleanSlice architecture, naming rules, and project structure.
Quick setup

Install in your editor

Click your editor below for copy-paste config.

Cursor
~/.cursor/mcp.json
+ enforcement rule
→
Claude Code
One command
+ enforcement rule + stop hook
→
VS Code
.vscode/mcp.json
+ enforcement rule
→
Windsurf
MCP config file
+ enforcement rule
→
Claude Desktop
claude_desktop_config.json
→
Opencode
Opencode config
+ enforcement rule
→
Endpoint https://mcp.cleanslice.org/mcp
Capabilities

What your AI gets access to

Four tools that turn any MCP-compatible agent into a CleanSlice expert.

⚡
get-started
Returns the essential CleanSlice rules and naming conventions. The AI calls this first to understand how to structure your project — slices, layers, file placement, and the critical singular-naming rule.
📂
list-categories
Lists all available documentation categories (quickstart, patterns, best practices, etc.) so the AI knows what deeper knowledge is available before diving in.
🔍
search
Full-text search across docs — filterable by query, framework (NestJS, Nuxt), development phase, feature name, category, or tags. Returns compact keyword-in-context snippets instead of full documents.
📄
read-doc
Read the full content of a specific document by path. Use after search when snippets aren't enough — returns the complete document with no truncation.
Advanced

Self-host or run locally

Clone the repo to run your own instance or customize documentation.

bash 
# Clone and run locally
git clone https://github.com/CleanSlice/mcp.git
cd mcp
npm install
npm run dev

# Or use Docker
docker build -t cleanslice-mcp .
docker run -p 8080:8080 cleanslice-mcp