mcp_procman

# MCP Process Manager A Python library for asynchronously launching, monitoring, and controlling external processes with comprehensive I/O management capabilities and logging. This allows your tool calling systems and Agents to run commands without the risk of blocking and halting progress waiting for a returned value that may never come. ## Features - Asynchronous process launching and monitoring - Combined stdout/stderr capture with configurable ring buffer - Thread-safe I/O operations with timeout support - Process interaction via stdin (lines and raw characters) - Output retrieval and searching capabilities (string, regex, wildcard) - Non-blocking design to prevent deadlocks - Comprehensive logging with rotation for diagnostics and debugging ## Installation ```bash # Clone the repository git clone git@github.com:NeuralNotwerk/mcp_procman.git # Install the package cd mcp_procman pip install -e . ``` ## Project Structure ``` mcp_process_manager/ ├── src/ │ └── mcp_process_manager/ │ ├── __init__.py # Package initialization │ ├── process_manager.py # Main ProcessManager class │ ├── process_handler.py # Individual process handler │ └── ring_buffer.py # Ring buffer implementation ├── test/ │ └── test_process_manager.py # Test suite ├── logs/ # Log directory (created automatically) ├── setup.py # Package setup script └── README.md # This file ``` ## Usage ### As a Library ```python from mcp_process_manager import ProcessManager # Create a process manager manager = ProcessManager() # Start a process status, pid = manager.process_start(["echo", "Hello, World!"]) # Get process output lines = manager.stdio_get_lines(pid) print(lines) # Start an interactive process status, pid = manager.process_start(["python", "-c", "while True: print(input('Enter text: '))"]) # Send input to the process manager.stdio_send_line(pid, "Hello from Python!") # Get process output lines = manager.stdio_get_lines(pid) print(lines) # Kill the process manager.process_kill(pid) # Remove the process from tracking manager.process_remove(pid) ``` ### As a Command-Line Tool After installation, you can run the MCP Process Manager as a command-line tool: ```bash # Run the MCP Process Manager mcp_procman ``` This will start the process manager with FastMCP integration, allowing you to interact with it via the command line. ## Logging The MCP Process Manager includes comprehensive logging capabilities: - Logs are stored in the `./logs/` directory (created automatically if it doesn't exist) - Log files follow the naming pattern `mcp_procman_YYYY_MM_DD.log` with daily rotation - Log files are limited to 10MB in size with rotation (up to 5 backup files) - All exceptions, function inputs, and function outputs are logged - Hierarchical loggers are used for different components: - `mcp_process_manager` (root logger) - `mcp_process_manager.process_manager` - `mcp_process_manager.process_handler` - `mcp_process_manager.ring_buffer` To access logs programmatically: ```python import logging # Get the logger logger = logging.getLogger("mcp_process_manager") # Set log level if needed logger.setLevel(logging.DEBUG) # Options: DEBUG, INFO, WARNING, ERROR, CRITICAL ``` ## API Reference ### Process Management - `process_start(command, timeout=15.0)`: Start a new process - `process_status(pid, timeout=15.0)`: Get process status - `process_kill(pid, timeout=15.0)`: Kill a process - `process_remove(pid, timeout=15.0)`: Remove a process from tracking - `process_list(timeout=15.0)`: List all tracked processes - `all_kill(timeout=15.0)`: Kill all running processes - `all_remove(timeout=15.0)`: Remove all non-running processes ### I/O Operations - `stdio_get_lines(pid, max_lines=5, timeout=15.0)`: Get recent output lines - `stdio_search_lines(pid, search_type, pattern, max_lines=5, timeout=15.0)`: Search process output - `all_search(search_type, pattern, max_lines_per_pid=2, timeout=15.0)`: Search across all processes - `stdio_send_line(pid, line, timeout=15.0)`: Send a line to process stdin - `stdio_send_chars(pid, chars, timeout=15.0)`: Send raw characters to process stdin ## Testing ```bash # Run tests python -m unittest discover -s test ``` ## License Apache 2.0