Author Steven Bucher shares the latest on AI Shell Preview 6, highlighting richer MCP integration, built-in tools for PowerShell, and improved error resolution features. Learn how these updates foster a more productive, AI-enhanced shell environment.

AI Shell Preview 6 Release!

By Steven Bucher

The AI Shell team is excited to announce the release of Preview 6, focusing on improved user experiences, robust error handling, tighter integration with Model Context Protocol (MCP) tools, and more automation capabilities inside your PowerShell workflows.

What’s New at a Glance

  • MCP client integration: AI Shell can now act as an MCP client, connecting to a variety of MCP servers for extended functionality.
  • Built-in tools: A suite of integrated commands enhances automation and enables richer context-awareness for the AI agent.
  • Resolve-Error improvements: The command is now smarter in identifying and resolving previous command errors.
  • Aliases and workflows: Stay productive within your terminal using intuitive shortcuts for AI Shell commands.

MCP Integration

AI Shell now works as an MCP client. This allows you to add any MCP server to boost AI Shell’s capabilities with server-specific tools and broader data access.

How to Add MCP Servers:

  • Create an mcp.json file in your $HOME\.aish\ folder.

  • Example configuration for two MCP servers (“everything” and “filesystem”):

    {
  "servers": {
    "everything":{
      "type":"stdio",
      "command":"npx",
      "args":["-y", "@modelcontextprotocol/server-everything"]
    },
    "filesystem": {
      "type": "stdio",
      "command": "npx",
      "args": [ "-y", "@modelcontextprotocol/server-filesystem", "C:/Users/username/" ]
    }
  }
}
  • For remote servers, set type to https.
  • View server registration in the AI Shell UI and verify with the /mcp command.
  • /mcp lists all MCP Servers and their available tools.

Requirements:

  • Node.js or uv should be installed for servers requiring them.

Standalone Experience:

  • Use servers like @simonb97/server-win-cli to run PowerShell, CMD, Git Bash, or other shell commands.
  • Define permitted commands and operations.

Note: Community MCP servers are not officially maintained by Microsoft; evaluate them before use. Additional servers can be found at:

Built-in Tools for AI Shell

A new lineup of built-in tools is now available to the AI agent, improving context-awareness and enabling new automations. These complement existing MCP Server tools, but are exclusive to the base AI Shell environment.

Tool Name Description
get_working_directory Returns the current working directory, provider name, and path of the PowerShell session.
get_command_history Fetches up to 5 of the latest commands executed in the session.
get_terminal_content Retrieves all terminal output currently displayed.
get_environment_variables Lists environment variables (with sensitive values redacted).
copy_text_to_clipboard Copies provided text/code to the system clipboard.
post_code_to_terminal Inserts code into the prompt for user review before execution.
run_command_in_terminal Executes shell commands in a persistent session, retaining context across commands.
get_command_output Retrieves output from commands launched via run_command_in_terminal.

Usage Notes: These tools work with a connected PowerShell session, providing detailed context for AI-driven interactions.

Demos:

  • Run shell commands via AI (run_command_in_terminal tool)
  • Use get_terminal_content for richer context to AI
  • Enhance agent responsiveness to session state changes

Resolve-Error Command Improvements

The Resolve-Error command is now more intelligent:

  • Identifies which command to troubleshoot based on error state and command history.
  • Smart detection based on $LastErrorCode, $?, and analysis of terminal content.
  • Reduces need for immediate AI intervention after errors; now, AI can effectively help resolve lingering or earlier issues.

Staying in Your Shell: Commands and Aliases

  • Invoke-AIShell: Main entry point, now with the askai alias.
  • Resolve-Error: Enhanced troubleshooting, now with the fixit alias.

This setup enables continued workflow within PowerShell without switching context, streamlining access to AI-powered assistance.

Command Name Alias
Invoke-AIShell askai
Resolve-Error fixit

Conclusion

This release is designed to empower users through better automation, more flexible error resolution, and contextually-aware AI assistance directly in PowerShell. For feedback and feature requests, please submit issues at AI Shell GitHub repository.

Thank you for your continued input and support!

AI Shell Team

Authors: Steven Bucher & Dongbo Wang

This post appeared first on “Microsoft PowerShell Blog”. Read the entire article here