AI Shell Preview 6 is here!

We are super excited to announce the latest preview release of AI Shell. This release focuses on enhancing the user experience with new features, improved error handling, and better integration with Model Context Protocol (MCP) tools.

What’s new at a glance

• MCP client integration
• Built-in tools
• Resolve-Error command improvements
• Aliases and flows for staying in your terminal

MCP Integration

AI Shell now acts as an MCP client, which allows you to add any MCP server to your AI Shell experience. Connecting to an MCP server massively improves the capability of your AI Shell giving you the tools that provide more relevant data or carry out actions!

Adding MCP Servers

To add an MCP server, create an mcp.json file in $HOME\.aish\ folder. The following example shows two MCP servers: everything and filesystem. You can add any MCP servers you want.

{
    "servers": {
      "everything":{
        "type":"stdio",
        "command":"npx",
        "args":["-y", "@modelcontextprotocol/server-everything"]
      },
      "filesystem": {
        "type": "stdio",
        "command": "npx",
        "args": [
          "-y",
          "@modelcontextprotocol/server-filesystem",
          "C:/Users/username/"
        ]
      }
    }
  }

If it’s a remote MCP server, change the type to https. You know that you have successfully added an MCP server when you see it in the AI Shell UI. You can confirm that it’s running by checking the status of the server through the /mcp command. Using /mcp also lists each MCP Server and the tools available.

Standalone experience with AI Shell and MCP Servers

MCP servers enhance your standalone experience with AI Shell, allowing your command line to use MCP servers and AI to perform tasks. For example, @simonb97/server-win-cli is an MCP server that allows you to run commands on your Windows machine, whether it be PowerShell, CMD, Git Bash, or any configured shell you use! It also provides configuration settings to define which commands and operations are allowed to run.

Additional MCP servers:

• DesktopCommander
• Other community MCP Servers

Built-in Tools for AI Shell

This release introduces built-in tools that are now accessible to agents within AI Shell. These commands are similar to MCP Server tools, but are exclusive to the AI Shell experience. These tools are designed to enhance the AI Shell experience by providing context-aware capabilities and automation features. They can be used in conjunction with the MCP servers to create a powerful AI-driven shell environment.

Here is a simple demo showing how you can have AI Shell run commands on your behalf using the run_command_in_terminal tool:

This example shows how additional context is provided to AI Shell to improve results:

You can also use the get_terminal_content tool to get the content from the connected terminal and provide it to AI Shell to help it understand what you are trying to do:

Resolve-Error Command Improvements

Previously the Resolve-Error command was only able to run after an error occurred in the previous command. Now, Resolve-Error identifies which command the user wants to troubleshoot:

• If the last error’s command matches the most recent command in history, it’s assumed to be the one the user is interested in.
• If the last error’s command isn’t the most recent and $LastErrorCode is null or zero, the error likely comes from an earlier command, not the very last one.
• If $LastErrorCode is non-zero and $? is false, the last command was a failing native command.
• If $LastErrorCode is non-zero but $? is true, it’s unclear which command or failure the user is focused on, so the agent analyzes the terminal content to determine the relevant context.

This logic allows AI Shell to better understand what the error the user is trying to resolve is rather than requiring you to ask for AI’s help immediately after an error occurs.

Staying in your shell

The Invoke-AIShell and Resolve-Error commands allow you to stay in your working terminal to interact with the AI Shell agent. To learn more about the parameters added, see the previous blog post that details these features. For your convenience, these commands have aliases that make them quicker to use.

Conclusion

We hope that these enhancements make your experience with AI Shell more powerful! We are always looking for feedback and suggestions, so please submit issues or feature requests in our GitHub repository.

Thank you so much!

AI Shell Team

Steven Bucher & Dongbo Wang