MCP Personal

1. File Search Server (mcp_fd_server.py)

Fuzzy file NAME search capabilities usingfdandfzf:

- Fast file name searchusingfd(searches file names/paths, NOT contents)
- Fuzzy filtering of file nameswithfzffor intelligent name matching
- Pattern matchingwith regex and glob support for file names
- Result limitingwithlimitparameter to restrict number of matches
- Proper error handlingfor fzf exit codes (distinguishes "no matches" from errors)
- Multiline mode(advanced): can also search file contents when enabled
- Standalone CLIfor testing and direct usage
- Key point: Primary purpose is finding files by NAME, not searching contents

2. Fuzzy Search Server (mcp_fuzzy_search.py)

Advanced search with both file name and content capabilities usingripgrepandfzf:

- File name fuzzy search- find files by partial/fuzzy names
- Content searchusingripgrepto search text within files
- Fuzzy filteringof results usingfzf --filter
- Proper error handlingfor fzf exit codes (distinguishes "no matches" from errors)
- Two distinct modes:

- fuzzy_search_files: Search file NAMES/paths
- fuzzy_search_content: Search file CONTENTS with path+content matching by default

- get_pdf_page_labels: Get all page labels from a PDF file
- get_pdf_page_count: Get the total number of pages in a PDF file
- get_pdf_outline: Extract table of contents/bookmarks from a PDF file

3. SQLite Server (mcp_sqlite_server.py)

SQLite database operations with configurable read/write permissions:

- Read-only by default- Write operations disabled unless explicitly enabled
- Agent-friendly- Clear tool descriptions and examples for easy AI agent usage
- In-memory database support- Use:memory:for temporary databases
- Comprehensive operations- Query, execute, list tables, describe schema, create tables
- Safety features- Query validation, write operation restrictions, clear error messages
- Standalone CLIfor testing and direct usage

4. Sequential Thinking Server (mcp_sequential_thinking.py)

# macOS/Linux curl -LsSf https://astral.sh/uv/install.sh | sh # Or using pip (if you already have Python) pip install uv

The PDF search and extraction tools in the Fuzzy Search Server areoptional. The server will work without these binaries installed - only the PDF-specific tools will be unavailable. This allows you to use the core fuzzy search functionality without requiring all dependencies.

The file search server requires the following command-line tools:

sudo apt install fd-find fzf # Note: On Debian/Ubuntu, fd is installed as 'fdfind'

brew install ripgrep fzf # For PDF search capabilities (optional) brew install ripgrep-all pandoc pip install PyMuPDF # Or: uv pip install PyMuPDF

sudo apt install ripgrep fzf # For PDF search capabilities (optional) # Install ripgrep-all cargo install ripgrep-all # Requires Rust/cargo # Install PyMuPDF pip install PyMuPDF # Or: uv pip install PyMuPDF # Install pandoc sudo apt install pandoc

git clone https://github.com/yourusername/mcp-personal.git cd mcp-personal

The easiest way to add MCP servers to Claude Code is using the CLI:

# Add custom Python servers from this repository # IMPORTANT: Use -s user for personal tools you want available across all projects # Without -s flag, servers are only available in current directory and are temporary # Easy method: Use $(pwd) when in the project directory cd /path/to/mcp-personal claude mcp add file-search -s user -- $(pwd)/mcp_fd_server.py claude mcp add fuzzy-search -s user -- $(pwd)/mcp_fuzzy_search.py claude mcp add sqlite -s user -- $(pwd)/mcp_sqlite_server.py claude mcp add sequential-thinking -s user -- $(pwd)/mcp_sequential_thinking.py # Or use relative paths (also from project directory) claude mcp add file-search -s user -- ./mcp_fd_server.py claude mcp add fuzzy-search -s user -- ./mcp_fuzzy_search.py claude mcp add sqlite -s user -- ./mcp_sqlite_server.py claude mcp add sequential-thinking -s user -- ./mcp_sequential_thinking.py # Using absolute paths (works from anywhere) claude mcp add file-search -s user -- /path/to/mcp-personal/mcp_fd_server.py claude mcp add fuzzy-search -s user -- /path/to/mcp-personal/mcp_fuzzy_search.py claude mcp add sqlite -s user -- /path/to/mcp-personal/mcp_sqlite_server.py claude mcp add sequential-thinking -s user -- /path/to/mcp-personal/mcp_sequential_thinking.py # Disable thought logging for the sequential thinking server claude mcp add sequential-thinking -s user -e DISABLE_THOUGHT_LOGGING=true -- /path/to/mcp-personal/mcp_sequential_thinking.py # Add SQLite server with write permissions enabled claude mcp add sqlite -s user -- /path/to/mcp-personal/mcp_sqlite_server.py --allow-writes # Or using environment variable claude mcp add sqlite -s user -e MCP_SQLITE_ALLOW_WRITES=true -- /path/to/mcp-personal/mcp_sqlite_server.py # Add Python servers with Python interpreter explicitly claude mcp add my-server -s user -- python /path/to/my_mcp_server.py # Add servers with arguments claude mcp add my-server -s user -- python /path/to/server.py arg1 arg2 # Add servers with environment variables claude mcp add my-server -s user -e API_KEY=your_key -e DEBUG=true -- python /path/to/server.py # Scope options: # -s local (default): Temporary, only in current directory # -s project: Shared with team via .mcp.json file # -s user: Personal, available across all your projects (recommended)

- The--separator is important before the command and its arguments
- Environment variables use-e KEY=valuesyntax
- Use-s userfor personal servers available across all projects
- Both relative and absolute paths work

For Claude Desktop, manually add servers to~/Library/Application Support/Claude/claude_desktop_config.json:

# Query database (read-only operations) ./mcp_sqlite_server.py query "SELECT  FROM users" database.db ./mcp_sqlite_server.py query "SELECT COUNT(*) as total FROM orders WHERE status = 'active'" sales.db # List all tables ./mcp_sqlite_server.py list-tables database.db # Describe table schema ./mcp_sqlite_server.py describe-table users database.db # Execute write operations (requires --allow-writes flag) ./mcp_sqlite_server.py execute "INSERT INTO users (name, email) VALUES ('John', 'john@example.com')" database.db --allow-writes ./mcp_sqlite_server.py execute "UPDATE users SET active = 0 WHERE last_login < date('now', '-1 year')" database.db --allow-writes # Use in-memory database for testing ./mcp_sqlite_server.py query "SELECT sqlite_version()" :memory:

Once configured, Claude can invokesequentialthinkingto work through problems in explicit, revisable steps. Natural-language prompts that trigger it:

- "Think step by step about how to design this caching layer"
- "Use sequential thinking to debug why this test is flaky"
- "Break this refactor down into thoughts and revise as you go"
- "Plan the migration with the sequential thinking tool, branching if you see alternatives"

Each call records one thought and returns the current state (thoughtNumber,totalThoughts,nextThoughtNeeded, activebranches, andthoughtHistoryLength). Claude keeps calling the tool — revising, branching, or extending the total as needed — untilnextThoughtNeededis false.

By default, formatted thoughts are pretty-printed tostderras they arrive (useful when running the server interactively or through the MCP Inspector). SetDISABLE_THOUGHT_LOGGING=truein the environment to silence the logging while keeping the structured tool output intact.

# Run with thought logging disabled DISABLE_THOUGHT_LOGGING=true ./mcp_sequential_thinking.py

Both MCP servers supportmultiline search modewhich changes the search behavior:

- File Search Server: Searches file NAMES/PATHS only
- Fuzzy Search Server: Searches line-by-line within file contents

- File Search Server: Switches to searching file CONTENTS instead of names
- Fuzzy Search Server: Treats entire file contents as single searchable units

Multiline mode is for searching file CONTENTS (not names):

- Finding class definitionswith their methods:"class UserService authenticate"(fuzzy match in contents)
- Locating function implementations:"async function await fetch"(all terms in one file)
- Searching for configuration blocks:"database host port"(finds files containing all terms)
- Finding code structuresacross lines:"try catch finally"(fuzzy matches across lines)
- Important: This searches CONTENTS, not file names!

File Search Server Multiline (Content Search)

# With --multiline, searches file CONTENTS instead of names # Find files CONTAINING these terms (not in file names) ./mcp_fd_server.py filter "class constructor method" "" src --multiline # Find Python files CONTAINING specific patterns ./mcp_fd_server.py filter "class def return" "" . --multiline # Find files CONTAINING database configuration ./mcp_fd_server.py filter "database host password" "" config --multiline

# search-files with --multiline searches file CONTENTS (not names) ./mcp_fuzzy_search.py search-files "async function await" src --multiline # search-content with --multiline treats files as single units ./mcp_fuzzy_search.py search-content "try catch finally" . --multiline # Find files CONTAINING class definitions with specific methods ./mcp_fuzzy_search.py search-content "class constructor render" src --multiline

- File size: Multiline mode reads entire files into memory; best for typical source code files
- Result size: Multiline results include complete file contents, which may be truncated for display
- Pattern complexity: Simple fuzzy patterns work well; complex queries may be slower
- Use specific terms:"class MyClass def method"is better than just"class def"(no regex!)
- Combine structure and content:"import React export default"finds React components
- Mind the output: Results show the entire matching file content
- Test incrementally: Start with simple patterns and refine

Both MCP servers usefzf's extended search syntaxfor powerful fuzzy filtering. Understanding this syntax will help you construct precise queries.

IMPORTANT: Thefuzzy_filterparameter infuzzy_search_contentdoes NOT support regular expressions. It uses fzf's fuzzy matching syntax as described below. If you need regex-like patterns, use the position anchors and exact matching features of fzf syntax instead.

Note: These examples show how to achieve regex-like filtering WITHOUT using regular expressions, sincefuzzy_filterdoes not support regex.

# Find Python configuration files, excluding tests config .py$ !test # Find main files in src directory with multiple extensions ^src/ main py$ | js$ | go$ # Find exact package manager files 'package.json' | 'yarn.lock' | 'Pipfile' # Find TODO comments in code files, excluding documentation TODO .py$ | .js$ | .go$ !README !docs/ # Find function definitions, excluding test files 'def ' .py$ !test !spec # Find configuration files with specific extensions, excluding backups config .json$ | .yaml$ | .toml$ !.bak$ !.old$

When usingfuzzy_search_content, queries work on the formatfile:line:content:

Default Mode (matches file paths AND content):

# Find implementation TODOs in specific file types TODO implement .py: | .js: # Matches TODO in .py or .js files # Find error handling in specific files error 'main.py:' | 'app.js:' # Matches 'error' in main.py or app.js # Find updates in test files test.py: update # Matches 'update' in files named test.py # Find async functions with error handling 'async def' error .py$ # Matches in Python files

# With --content-only flag or content_only=true parameter # These will ONLY match the content, not file names: # Find TODO comments regardless of filename TODO implement # Won't match files named 'TODO.txt' # Find async/await patterns async await catch # Pure content search # Find class definitions 'class ' 'def __init__' # Won't match 'class.py' filename