Learn how to use ClaudeBoost

Getting Started

ClaudeBoost is a free MCP (Model Context Protocol) plugin for Claude Code that automatically enhances your prompts before they reach Claude. Install it in 30 seconds.

Step 1: Install & configure (regular terminal)

pip install claudeboost-mcp
claudeboost-mcp --setup

Setup asks for your Anthropic API key, installs slash commands, and offers to sign in (for dashboard sync).

Step 2: Register MCP server (inside Claude Code)

Open Claude Code and type:

/boost --setup

This registers the MCP server. You only need to do this once. Restart Claude Code after this step.

Step 3: Start using it

That's it. Auto-boost is on by default — every task prompt you type is automatically enhanced. You don't need to type /boost every time, just type normally.

> build me a REST API for user management
# ⚡ CLAUDEBOOST · general_coding · Score: 8/30 → 22/30 (+14)
# 🔧 Boost added: tests & success criteria, organized sections
# Choose: Use boosted / Add notes / Keep original

How It Works

When you type a prompt in Claude Code, ClaudeBoost intercepts it and runs a 4-step pipeline before Claude sees your instruction:

Commands

ClaudeBoost adds several slash commands and modifiers to Claude Code:

Terminal Commands

Run these in your regular terminal (outside Claude Code):

Claude Code Commands

Run these inside Claude Code:

Examples

# First time only — register the MCP server
> /boost --setup
✅ MCP server registered. Restart Claude Code to activate.

# After restart, auto-boost is on — just type normally
> build me an API endpoint for user auth
⚡ CLAUDEBOOST · general_coding · Score: 8/30 → 22/30 (+14)
🔧 Boost added: tests & success criteria, organized sections

# Skip boost for one prompt
> fix this typo in readme.md --raw

# Sign in (email + password in terminal)
> /boost --login

# Check which account is connected
> /boost --status

# Change boost level
> /boost-settings -l full

# Well-structured prompts skip automatically
> /boost fix the 403 error in src/auth/session.ts by adding token refresh
✅ Your prompt scores 22/30 — already well-structured!

Boost Levels

ClaudeBoost offers three intensity levels. Each targets a different quality score threshold:

Domains

ClaudeBoost auto-classifies every prompt into one of 7 domains using Claude Haiku. Each domain has tailored enhancement rules:

Scoring System

Every prompt is scored 1-5 on six dimensions, for a total of 6-30. Scoring is automated via text analysis — no API calls required.

Quality Levels

Feedback & RLHF

ClaudeBoost learns from your preferences through a feedback loop inspired by Reinforcement Learning from Human Feedback (RLHF).

When you rate a boost (1-5 stars) or leave text feedback, it's stored in~/.claudeboost/history.json. Next time ClaudeBoost enhances a prompt in the same domain, it loads your last 5 feedback entries + your domain constraints and injects them into the enhancement prompt. The more feedback you give, the more personalized your boosts become.

Constraints

Domain constraints are persistent rules that ClaudeBoost always applies when enhancing prompts in a specific domain. Set them in the Dashboard under Constraints, or via the CLI settings.

Quick Presets

The Dashboard includes 12 one-click presets for common stacks: Python Data Science, PyTorch ML, dbt+BigQuery, Spark Pipeline, React+TypeScript, Next.js App Router, FastAPI, AWS+Terraform, Docker+Kubernetes, GitHub Actions CI/CD, API Documentation, and Executive Reports. Click one to fill in the domain constraints instantly.

Custom Constraints

Dashboard

ClaudeBoost includes a web dashboard with authentication. Sign up at the landing page and your data syncs between CLI and web. The dashboard auto-refreshes every 5 seconds.

Pages

Running the Dashboard

cd web-dashboard
npm install
npm run dev
# Dashboard available at https://claudeboost.vercel.app

FAQ