Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
This page provides help with common questions about AI Shell, also known as Project Mercury.
What is Project Mercury?
Project Mercury was the code name for AI Shell and is a platform that provides a framework for developers to build their own AI Agents and assistance providers for an AI Shell. Agents provide the user experience for the LLM and are deeply connected to PowerShell 7. For more, see the AI Shell architecture.
What operating systems are supported?
We've tested on macOS and Windows operating systems. AI Shell may work on linux but we haven't tested it. We can't guarantee that all features will work as expected.
What are the different ways I can use AI Shell?
You can use AI Shell in two ways:
- As a standalone application
- As a split pane experience in PowerShell 7 and Windows Terminal
The split pane experience is based on the capabilities of your terminal. terminal. For example, you
can split the Windows Terminal pane by running the following command: wt -w 0 sp. Refer to the
documentation for your terminal application to see if it supports this feature.
Note
Not all terminal applications support this feature.
We've made this experience easier when using PowerShell 7 and Windows Terminal. For more information see Get started with AI Shell in PowerShell.
What is the difference between the standalone application and the split pane experience?
The split pane experience with PowerShell 7 has deeper integration with PowerShell and thus can communicate across the panes for inserting AI generated codes to the working shell and sending errored commands to the AI pane for assistance. The standalone application is a more generic experience that can be used with any shell and has no such integration.
What are AI agents?
An agent is a code library that implements the interfaces that talk to a specific language model or assistance provider. An assistance provider is an agent that provides user assistance without using a language model or AI engine.
We currently support the following agents:
Users interact with these agents using natural language in a conversational manner.
How do I create an agent?
To create an agent, you need to implement the IAgent interface. For more information, see
Create an Ollama Agent for the details of creating an agent. You can use this agent as a
template to create your own agent.
You can use the example code as written to interact with Ollama, or you can configure the OpenAI agent to interact with the Ollama using the OpenAI-compatible endpoint.
How do I share my agent or find other community built agents?
There isn't a centralized repository for sharing community built agents. In the interim, we recommend sharing it on the discussions page of the Project Mercury repository.