better-finder-mcp

GitHamza0206/better-finder-mcp

3.3

If you are the rightful owner of better-finder-mcp and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to henry@mcphub.com.

Better Finder MCP is a tool designed for intelligent file discovery using a Git-like workflow, integrating semantic search and Claude MCP for enhanced file management.

🔍 Better Finder MCP

Git-like workflow for intelligent file search and Claude MCP integration

Better Finder transforms file discovery with semantic search, hybrid matching, and a familiar Git-style workflow. Index your documents, search with natural language, and integrate seamlessly with Claude Desktop.

🚀 Quick Start

Installation

# Clone and install
git clone https://github.com/GitHamza0206/better-finder-mcp.git
cd better-finder-mcp
uv venv
source .venv/bin/activate
pip install -e .

# Install it globally with uv
uv tool install -e .

Basic Workflow

# 1. Stage files for indexing (like git add)
better-finder add ~/Documents

# 2. Check what's staged
better-finder status

# 3. Index staged files
better-finder index

# 4. Search your files
better-finder search "quarterly financial reports"

📋 Commands

File Staging

  • better-finder add <path> - Stage files or Documents for indexing
  • better-finder rm <path> - Remove files from staging
  • better-finder status - Show staged files
  • better-finder index - Index staged files

Search & Management

  • better-finder search <query> - Search indexed files
  • better-finder stats - Show index statistics
  • better-finder clear-index - Clear all indexed data
  • better-finder server - Start MCP server for Claude

Utilities

  • better-finder show <file> - Display file content
  • better-finder remove-file <file> - Remove file from index

🎯 Key Features

Hybrid Search

Combines semantic search, keyword matching, and fuzzy filename search for comprehensive results.

Git-like Workflow

Familiar staging process gives you precise control over what gets indexed.

.betterfinderignore Support

Use ignore patterns to exclude sensitive files:

# .betterfinderignore
secrets/
*.key
temp-*.pdf
node_modules/

Supported File Types

  • Documents: PDF, DOC, DOCX, TXT, MD, RTF, ODT
  • Spreadsheets: XLSX, XLS, CSV, ODS
  • Presentations: PPTX, PPT
  • Data: JSON, XML

Claude MCP Integration

Start the MCP server to use Better Finder directly within Claude Desktop:

better-finder server

Add to your Claude Desktop config (~/.cursor/mcp.json):

{
  "mcpServers": {
    "better-finder": {
      "command": "better-finder",
      "args": ["server"]
    }
  }
}

🔧 Configuration

Better Finder uses smart defaults:

  • Default scan paths: ~/Documents, ~/Desktop, ~/Downloads
  • Ignored directories: .git, node_modules, .venv, cache folders
  • Chunk size: 2000 characters for better context
  • Search threshold: 0.4 for comprehensive results

💡 Examples

Document Discovery

# Find presentations about sales
better-finder search "sales presentation Q4"

# Look for specific file types
better-finder search "budget" --type excel

# Get more results
better-finder search "meeting notes" --max 20

Staging Workflow

# Stage entire project documentation
better-finder add ./docs

# Remove sensitive files
better-finder rm ./docs/secrets/

# Check what will be indexed
better-finder status

# Index everything staged
better-finder index

MCP Integration

Once the server is running, ask Claude:

  • "Search my documents for budget reports"
  • "Find presentations about project timelines"
  • "Show me files related to client proposals"

🏗️ Architecture

  • File Processors: Extract content from different formats
  • FAISS Vector Store: Fast similarity search with sentence transformers
  • SQLite Metadata: Efficient file metadata and chunk storage
  • Staging System: Git-like file management with JSON persistence
  • MCP Server: Model Context Protocol integration for Claude

📊 Performance

  • Indexing: ~100-500 files per minute
  • Search: Sub-second results
  • Memory: ~200-500 MB for typical collections
  • Storage: ~1-5 MB per 1000 documents

📝 License

MIT License - see LICENSE file for details.


Made for developers who want intelligent file discovery with familiar Git-like controls.