systerd-lite

You run systerd-lite to empower an AI-native agent that can monitor, control, and optimize a Linux system through an MCP server. It exposes a broad set of system, security, container, and self-editing capabilities, and provides a browser-based UI for management while supporting programmatic access via HTTP. This guide shows you how to use, install, and configure the MCP server so you can safely automate system operations with an LLM.

How to use

Start the MCP server and connect with an MCP client to access the full range of tools. You can call monitoring, security, system, container, self-editing, calculator, and config tools to inspect the host, manage services, run containers, edit the workspace, or apply templates. Use the Gradio UI for interactive management in your browser, or send requests through the HTTP API for automated workflows. You’ll typically begin by verifying the health endpoint, then start issuing tool calls to retrieve system information and perform actions as needed.

How to install

Prerequisites: you should have Git and Python available on your Linux system. You may also want a browser to access the Gradio UI.

Step 1: Clone the project repository.

Step 2: Change into the project directory.

Step 3: Make the startup script executable and start the MCP server.

Step 4: Open the Gradio UI in your browser at the provided URL and verify the HTTP API health endpoint.

Endpoints and startup options

The MCP server exposes an HTTP API at http://localhost:8089, a Gradio UI at http://localhost:7861, and a health check at http://localhost:8089/health. You can start the server with the recommended startup script, or run the main application with optional flags for ports, UI, headless mode, and debugging.

# Quick start (example)
# Clone and enter project
git clone https://github.com/your/sisterd_lite.git
cd sisterd_lite

# Start (dependencies auto-installed)
chmod +x start-mcp.sh
./start-mcp.sh

Configuration and security

The server supports multiple tool categories with per-tool permissions and templates. You can apply predefined templates to enable groups of tools, adjust permission levels, and manage access via a permission matrix. Recommended practice for production is to use a safe template (minimal or monitoring) and limit access to the HTTP API with appropriate network controls.

Permission levels include DISABLED, READ_ONLY, AI_ASK, and AI_AUTO. Use AI_ASK or AI_AUTO only in trusted environments, and consider enabling the self-edit tools only in controlled scenarios.

Template examples are available to quickly configure tool sets. For instance, the development template enables a broad set of tools including self-edit capabilities, while the minimal template focuses on safe monitoring.

Tools overview

Systerd-lite provides a comprehensive toolkit to monitor and manage the Linux system, as well as to let an LLM edit its own environment. The following tools are described in the available capabilities:

Monitoring: get_system_info, get_cpu_info, get_memory_info, get_disk_usage, get_temperature, list_processes, get_top_processes.

Security: get_selinux_status, get_apparmor_status, list_firewall_rules, scan_suid_files, get_failed_logins, audit_permissions.

System: manage_service, list_units, get_kernel_modules, get_hardware_info, get_usb_devices.

Container: list_containers, start_container, stop_container, run_container, get_container_logs.

Self editing: read_workspace_file, write_workspace_file, append_to_file, list_workspace_directory, search_workspace, execute_shell_command, install_python_package, get_python_environment, set_environment_variable, restart_self, get_self_status, backup_workspace.

Calculator: calculate, convert_units, matrix_operation, statistics, solve_equation.

MCP Config: get_mcp_config, list_mcp_tools, set_mcp_tool_permission, apply_mcp_template, get_mcp_templates.

Example client setup

Connect to the HTTP endpoint from your MCP client and begin calling tools such as get_system_info or get_disk_usage. You can also switch to the Gradio UI for interactive exploration of tools and permissions.

Notes

If you want to customize the environment, you can adjust tool permissions and apply templates to suit your security posture and workflow needs. Always verify the health endpoint after starting the server and use the UI to explore available tools and configurations.

Available tools

get_system_info

Provides a system overview including general hardware and software state.

get_cpu_info

Returns detailed CPU information such as cores, speed, and utilization.

get_memory_info

Reports current memory usage and availability.

get_disk_usage

Shows disk usage statistics for mounted filesystems.

get_temperature

Reads temperature sensors to monitor thermal status.

list_processes

Lists currently running processes.

get_top_processes

Identifies top processes by resource usage.

get_selinux_status

Returns the status of SELinux on the host.

get_apparmor_status

Returns the status of AppArmor on the host.

list_firewall_rules

Lists active firewall rules.

scan_suid_files

Finds SUID/SGID files for security auditing.

get_failed_logins

Reports failed login attempts.

audit_permissions

Audits file and directory permissions.

manage_service

Controls systemd services (start, stop, restart, enable).

list_units

Lists systemd units present on the system.

get_kernel_modules

Reports loaded kernel modules.

get_hardware_info

Provides hardware details for the system.

get_usb_devices

Lists connected USB devices.

list_containers

Lists running containers (e.g., Docker/alternative runtimes).

start_container

Starts a specified container.

stop_container

Stops a specified container.

run_container

Runs a new container from an image.

get_container_logs

Retrieves logs from a container.

read_workspace_file

Reads a file from the LLM workspace.

write_workspace_file

Writes or creates a file in the workspace.

append_to_file

Appends content to an existing workspace file.

list_workspace_directory

Lists contents of a workspace directory.

search_workspace

Searches for files or content within the workspace.

execute_shell_command

Executes a shell command within the workspace environment.

install_python_package

Installs a Python package into the workspace environment.

get_python_environment

Retrieves information about the Python environment.

set_environment_variable

Sets an environment variable for the workspace.

restart_self

Restarts the MCP server itself.

get_self_status

Reports the current status of the MCP server.

backup_workspace

Backs up the workspace contents.

calculate

Evaluates mathematical expressions.

convert_units

Converts between different measurement units.

matrix_operation

Performs matrix calculations.

statistics

Computes basic statistics.

solve_equation

Solves algebraic equations.

get_mcp_config

Retrieves current MCP configuration.

list_mcp_tools

Lists all available MCP tools.

set_mcp_tool_permission

Sets permissions for individual tools.

apply_mcp_template

Applies a predefined template of tools.

get_mcp_templates

Lists available MCP templates.