Booster

MCP Python License: MIT

🎯 What You Get

Booster MCP turns complex codebases into understandable systems.

You spend less time searching and more time building. Instead of:

• ❌ Manually grepping through thousands of files
• ❌ Getting lost in unfamiliar architectures
• ❌ Debugging by guessing

You get:

• ✅ Instant semantic understanding — ask questions, get answers across your entire codebase
• ✅ Visual architecture maps — see your code as a 3D "Code City"
• ✅ Automatic context injection — include exactly what the AI needs to fix bugs or add features
• ✅ Debugging superpowers — flipchart sessions with call graphs, sequence diagrams, and session notes
• ✅ 7 battle-tested agent skills — pre-built workflows for onboarding, refactoring, bug hunting, and code review

Real-World Impact

Problem	Old Way	With Booster
New developer onboarding	2-3 days of grep	30 min interactive map + context
Finding related code	Manual search	Semantic search + symbol graph
Debugging production issues	Stack traces → manual tracing	analyze_error() → call graph → fix
Code review bottlenecks	Hours of manual review	Semantic + dependency analysis

---

⚡ Quick Start (1 minute)

One-Click Install

macOS / Linux:

curl -fsSL https://raw.githubusercontent.com/NeuroGhostDev/Booster-mcp/main/install.sh | bash

Windows (PowerShell):

Invoke-WebRequest -Uri "https://raw.githubusercontent.com/NeuroGhostDev/Booster-mcp/main/install.ps1" -OutFile "install.ps1" | .\install.ps1

First Command

add_repo("C:\\my-project")
get_code_city()  # Opens 3D visualization in browser

That's it. Your codebase is now AI-searchable.

---

💎 Features at a Glance

🔍 Semantic Search (Not Just Grep)

• Vector-powered code search understands intent, not just keywords
• Find "authentication flow" and get relevant login, OAuth, token validation code
• Works across multiple languages

🗺️ Code City 3D

• Visual representation of your project structure
• Buildings = files (height = complexity, color = language)
• Instantly spot architectural patterns and dependencies

📊 Repository Maps

• One-page project overview for onboarding
• Auto-generated, always in sync
• Includes conventions, stack, and module boundaries

🐛 Flipchart Debugging

• Create debug sessions with call graphs and sequence diagrams
• Mermaid-powered visualizations
• Add notes and track hypotheses across symbols

🤖 7 Built-in Agent Skills

• booster-onboard — new codebase? Start here
• booster-context-inject — give AI exactly what it needs
• booster-bug-hunt — stack trace → diagnosis → fix
• booster-feature-add — find patterns, add consistent code
• booster-deep-dive — understand architecture
• booster-refactor — impact analysis + change automation
• booster-review — semantic code review

🔗 Context Injection

repo://map          → Repository structure
repo://stack        → Dependencies and frameworks
repo://conventions  → Code patterns and standards

Auto-sync with live documentation via fetch_stack_docs().

⚙️ Incremental Indexing

• Watchdog monitors changes
• No manual reindex needed
• Multi-repo support without server restart

---

📦 What's Included

Component	Purpose
FastMCP Server	Core tool definitions and routing
Vector Indexer	FAISS-backed semantic search across code
Graph Engine	Call graphs, import graphs, dependency analysis
Code City Visualizer	3D HTML visualization of codebase
Agent Skills	7 pre-built workflows for common tasks
Web UI	Manage repos, generate visualizations, inspect maps
Flipchart	Debug sessions with Mermaid diagrams
Toolkit	Grep, git, command execution, error analysis

---

🛠️ Installation

Requirements

• Python 3.11+
• Git
• Internet (first run downloads embedding model)

Full Manual Install (all platforms)

Clone & Setup:

git clone https://github.com/NeuroGhostDev/Booster-mcp.git
cd Booster-mcp
python3.11 -m venv .venv
source .venv/bin/activate  # or .\.venv\Scripts\Activate.ps1 on Windows
pip install -r requirements.txt

Start:

python server.py

Configure Your MCP Client

Add to your .claude_desktop_config.json (Claude Desktop, Cline, etc):

macOS/Linux:

{
  "mcpServers": {
    "Booster": {
      "command": "/home/user/Booster-mcp/.venv/bin/python",
      "args": ["/home/user/Booster-mcp/server.py"]
    }
  }
}

Windows (use absolute paths):

{
  "mcpServers": {
    "Booster": {
      "command": "C:\\Users\\YourName\\Booster-mcp\\.venv\\Scripts\\python.exe",
      "args": ["C:\\Users\\YourName\\Booster-mcp\\server.py"]
    }
  }
}

---

🚀 Common Workflows

Onboard a New Project

add_repo("C:\\project")
repo_stats()           # Quick metrics
get_repo_map()         # Architecture overview
semantic_search("main entry point")
find_symbol("main")    # Navigate to key functions

Debug a Production Error

analyze_error("TypeError: 'NoneType' object is not subscriptable")
flipchart_quick_debug("handler_function", max_depth=3)
read_with_context("auth.py", line=42, context=10)
git_diff("auth.py")    # See recent changes

Add a Feature Safely

semantic_search("similar feature pattern")
find_symbol("existing_feature")
flipchart_sequence_diagram("existing_feature", depth=5)
# Now safe to implement

Code Review with Superpowers

flipchart_call_graph("modified_function", max_depth=5)
external_deps("modified_function")
find_duplicates(min_lines=5)

---

📚 Full Documentation

• COOKBOOK.md — Deep-dive recipes and advanced usage
• MARKETPLACE.md — Publish to MCP catalogs

---

🎓 System Instructions for Maximum Power

Add this to your MCP client for best results:

MCP Usage Policy:
- New codebase? Use onboard skill first
- Production error? Use bug-hunt skill
- Architecture question? Use deep-dive skill
- Adding feature? Use feature-add skill + inject context
- Refactoring? Use refactor skill + analyze impact
- Code review? Use review skill + semantic analysis

Always prefer semantic tools over grep for understanding.

[See full system instructions in README]

---

🛠️ All Available Tools

Repository Management

• add_repo(path) — Index a new repository
• remove_repo(path) — Stop tracking
• reindex_repo(path) — Force re-index
• list_repos() — See active repos
• repo_stats() — Size, symbols, metrics

Search & Navigation

• semantic_search(query) — Find code by meaning
• find_symbol(name) — Locate functions, classes

Context Injection

• inject_context() — Auto-build AI context
• fetch_stack_docs() — Live dependency docs

Debugging

• flipchart_quick_debug(symbol) — Instant graphs
• flipchart_call_graph(symbol) — Show callers/callees
• flipchart_sequence_diagram(symbol) — Flow diagram
• analyze_error(stacktrace) — Error analysis

Utilities

• code_grep(pattern) — Smart grep
• read_with_context(file, line) — Show code + context
• git_log(path) — See history
• run_command(cmd) — Execute tools
• find_duplicates() — Code duplication
• external_deps(symbol) — See dependencies

Visualization

• get_code_city() — 3D visualization
• get_repo_map() — Architecture map

---

🌍 Language Support

• Python
• JavaScript / TypeScript
• Rust
• Go
• Java
• C / C++

---

🧪 Testing

python test_mcp.py        # Basic tests
python test_all.py        # Full suite

---

❓ FAQ

Q: Will this slow down my codebase?
A: No. Indexing is separate. Your code runs normally.

Q: Does it work with private repos?
A: Yes. Everything runs locally. No data leaves your machine.

Q: Which AI clients are supported?
A: Any MCP-compatible client (Claude Desktop, Cline, Continue, etc).

Q: Can I index multiple repos at once?
A: Yes. add_repo() as many as you want. No restart needed.

Q: What if my repo is 1M+ lines?
A: Works fine. Semantic indexing is built for scale. Use .ignore to skip heavy directories.

---

📖 Detailed Docs & Guides

• COOKBOOK.md — Recipes, examples, advanced techniques
• MARKETPLACE.md — Publish to Smithery, Glama

---

🤝 Contributing

1. Fork it
2. Branch: git checkout -b feature/my-feature
3. Commit: git commit -m "Add feature"
4. Test: python test_all.py
5. Push & open PR

---

📄 License

MIT — Use freely, commercial or otherwise.

---

🎯 Next Steps

1. Install now: Run the one-click installer
2. Try it: add_repo() + get_code_city()
3. Read cookbook: Advanced workflows and examples
4. Share feedback: GitHub Issues

---

Made for developers who demand more from their tools. ⚡