About Autonomous Shell Plugin

Overview

The Autonomous Shell Plugin empowers AI agents with the ability to interact directly with your system's shell. It provides a robust set of tools for session management, command execution, and real-time monitoring, all while maintaining a strong focus on security and efficiency.

Persistent Sessions

Maintain state across multiple commands by creating dedicated shell environments.

Async Capabilities

Execute long-running tasks in the background without blocking the AI's reasoning process.

MCP Tools Reference

Session

create_shell_session

Initializes a new shell environment. Returns a sessionId required for subsequent commands.

Execution

execute_command_sync

Runs a command and waits for completion. Best for quick operations like ls or cat.

Execution

execute_command_async

Starts a background process. Returns a commandId to track progress via get_command_result.

Monitoring

get_command_result

Fetches output from an async command. Uses Token Optimization if enabled to minimize data transfer.

Configuration Logic

Safety Filter Mechanism

The safety filter intercepts every command before it reaches the shell. It uses the following logic:

if (mode == "whitelist") {
    return commandList.contains(baseCommand);
} else {
    return !commandList.contains(baseCommand);
}
                        

Token Optimization

When enabled, the plugin tracks the last returned output for each command. Subsequent calls to get_command_result only return the delta (newly generated text), significantly reducing token consumption for long-running logs.

Security Warning

This plugin grants direct access to your system's shell. While the safety filter provides a layer of protection, it is not a substitute for careful monitoring. Never run the editor with elevated privileges (root/admin) unless absolutely necessary, and always review the command list configuration.