We’re tremendous excited to announce the most recent preview launch of AI Shell. This launch focuses on
enhancing the person expertise with new options, improved error dealing with, and higher integration
with Mannequin Context Protocol (MCP) instruments.
What’s new at a look
- MCP shopper integration
- Constructed-in instruments
Resolve-Errorcommand enhancements- Aliases and flows for staying in your terminal
MCP Integration
AI Shell now acts as an MCP shopper, which lets you add any MCP server to your AI Shell
expertise. Connecting to an MCP server massively improves the aptitude of your AI Shell giving
you the instruments that present extra related information or perform actions!
Including MCP Servers
So as to add an MCP server, create an mcp.json file in $HOME.aish folder. The next instance
reveals two MCP servers: all the things and filesystem. You’ll be able to add any MCP servers you need.
{
"servers": {
"all the things":{
"sort":"stdio",
"command":"npx",
"args":["-y", "@modelcontextprotocol/server-everything"]
},
"filesystem": {
"sort": "stdio",
"command": "npx",
"args": [
"-y",
"@modelcontextprotocol/server-filesystem",
"C:/Users/username/"
]
}
}
}If it’s a distant MCP server, change the sort to https. that you’ve got efficiently added
an MCP server while you see it within the AI Shell UI. You’ll be able to verify that it’s operating by checking the
standing of the server by means of the /mcp command. Utilizing /mcp additionally lists every MCP Server and the
instruments out there.
NOTE
You should have Node.js or uv put in to make use of MCP servers that
use these command strains instruments.
Standalone expertise with AI Shell and MCP Servers
MCP servers improve your standalone expertise with AI Shell, permitting your command line to make use of MCP
servers and AI to carry out duties. For instance, @simonb97/server-win-cli is an MCP server that
lets you run instructions in your Home windows machine, whether or not it’s PowerShell, CMD, Git Bash, or any
configured shell you employ! It additionally gives configuration settings to outline which instructions and
operations are allowed to run.
CAUTION
Please observe this can be a neighborhood MCP server and never an
official Microsoft MCP Server. We encourage you to do your individual analysis and testing earlier than utilizing
it.
Further MCP servers:
Constructed-in Instruments for AI Shell
This launch introduces built-in instruments that at the moment are accessible to brokers inside AI Shell. These
instructions are just like MCP Server instruments, however are unique to the AI Shell expertise. These instruments
are designed to boost the AI Shell expertise by offering context-aware capabilities and
automation options. They can be utilized at the side of the MCP servers to create a strong
AI-driven shell setting.
| Instrument Identify | Description |
|---|---|
get_working_directory |
Get the present working listing of the linked PowerShell session, together with the supplier identify (e.g., FileSystem, Certificates) and the trail (e.g., C:, cert:). |
get_command_history |
Rise up to five of the newest instructions executed within the linked PowerShell session. |
get_terminal_content |
Get all output at present displayed within the terminal window of the linked PowerShell session. |
get_environment_variables |
Get setting variables and their values from the linked PowerShell session. Values of doubtless delicate variables are redacted. |
copy_text_to_clipboard |
Copy the supplied textual content or code to the system clipboard, making it out there for pasting elsewhere. |
post_code_to_terminal |
Insert code into the immediate of the linked PowerShell session with out executing it. The person can assessment and select to run it manually by urgent Enter. |
run_command_in_terminal |
This software lets you execute shell instructions in a persistent PowerShell session, preserving setting variables, working listing, and different context throughout a number of instructions. |
get_command_output |
Get the output of a command beforehand began with run_command_in_terminal. |
Observe
The built-in instruments depend on the side-car expertise with a
linked PowerShell session and supply enhanced context consciousness and automation capabilities.
Right here is a straightforward demo displaying how one can have AI Shell run instructions in your behalf utilizing the
run_command_in_terminal software:
This instance reveals how extra context is supplied to AI Shell to enhance outcomes:
You can even use the get_terminal_content software to get the content material from the linked terminal and
present it to AI Shell to assist it perceive what you are attempting to do:
Resolve-Error Command Enhancements
Beforehand the Resolve-Error command was solely capable of run after an error occurred within the earlier
command. Now, Resolve-Error identifies which command the person desires to troubleshoot:
- If the final error’s command matches the newest command in historical past, it’s assumed to be the one
the person is involved in. - If the final error’s command isn’t the newest and
$LastErrorCodeis null or zero, the error
possible comes from an earlier command, not the final one. - If
$LastErrorCodeis non-zero and$?is fake, the final command was a failing native command. - If
$LastErrorCodeis non-zero however$?is true, it’s unclear which command or failure the person
is targeted on, so the agent analyzes the terminal content material to find out the related context.
This logic permits AI Shell to raised perceive what the error the person is attempting to resolve is
fairly than requiring you to ask for AI’s assist instantly after an error happens.
Staying in your shell
The Invoke-AIShell and Resolve-Error instructions help you keep in your working terminal to
work together with the AI Shell agent. To study extra concerning the parameters added, see the
earlier weblog put up that particulars these options. On your comfort, these instructions have
aliases that make them faster to make use of.
| Command Identify | Alias |
|---|---|
Invoke-AIShell |
askai |
Resolve-Error |
fixit |
set up AI Shell
To put in the most recent model of AI Shell, run the next command in your PowerShell terminal:
Invoke-Expression "& { $(Invoke-RestMethod 'https://aka.ms/install-aishell.ps1') }"We hope that these enhancements make your expertise with AI Shell extra highly effective! We’re all the time
in search of suggestions and recommendations, so please submit points or function requests in our
GitHub repository.
Thanks a lot!
AI Shell Workforce
Steven Bucher & Dongbo Wang
