Built In MCP Tools

PowerShell Universal includes built-in MCP tools at /api/v1/mcp. They help MCP clients inspect the PowerShell Universal configuration repository, discover cmdlets, read job output and notifications, and apply supported repository changes.

Built-in tools are separate from custom AI Tools. Connect an MCP client to /api/v1/mcp and authenticate with a PowerShell Universal app token.

Connect from Visual Studio Code

GitHub Copilot can call the built-in tools when Visual Studio Code is connected to the PowerShell Universal MCP server.

Press Ctrl+Shift+P, run MCP: Add Server..., select HTTP, and enter the MCP endpoint URL. By default, the URL is http://localhost:5000/api/v1/mcp.

The resulting settings.json configuration looks like this:

"mcp": {
  "servers": {
    "PSU": {
      "url": "http://localhost:5000/api/v1/mcp"
    }
  }
}

To authenticate with a PowerShell Universal app token, include it as a bearer token in the Authorization header:

"mcp": {
  "servers": {
    "PSU": {
      "url": "http://localhost:5000/api/v1/mcp",
      "headers": {
        "Authorization": "Bearer <your-app-token>"
      }
    }
  }
}

Replace <your-app-token> with a token for the PowerShell Universal user and roles the client should use. Do not commit a configuration containing a real app token to source control.

If the connection succeeds, GitHub Copilot displays the available PowerShell Universal tools.

Access control

Most built-in tools require the Administrator or Operator role. search_ant_design_controls also supports the App Editor role. get_job_output_by_id requires an authenticated user and verifies that user has permission to read the requested job or its configured read permission.

Use a dedicated app token with only the role required by the MCP client. Repository write tools can create, update, move, and delete configuration files, so do not grant Administrator or Operator access to an untrusted MCP client.

Typical flow

1. Call list_resource_types and list_resources to identify the configuration resource.
2. Use the repository and command tools to inspect the relevant files and cmdlets.
3. Apply a deliberate repository change with a write tool when authorized.
4. Call reload_resource to reload the affected resource.
5. Use get_notifications or get_job_output_by_id to verify the result.

Resource and command tools

Tool	Description
list_resource_types	Lists available PowerShell Universal resource types.
list_resources	Lists resources of a selected type, with wildcard filtering.
reload_resource	Reloads a selected resource type after a configuration change.
list_commands	Lists commands from a PowerShell module.
command_help	Returns focused or full help for a command, parameters, and examples.
list_app_commands	Lists Universal App-development cmdlets.

Repository and operational tools

Repository tools

The following tools inspect and manage repository-relative files and directories:

Tool	Description
list_files	Lists files and directories.
search_files	Searches file and directory names.
search_file_content	Searches repository file contents.
get_file_content	Reads a repository file.
set_file_content	Replaces a repository file's content.
new_file_system_item	Creates a repository file or directory.
move_file_system_item	Moves or renames a repository item.
delete_file_system_item	Deletes a repository item.

All paths are repository-relative. Use the discovery tools before changing a file, and call reload_resource after an authorized configuration change.

Operational tools

Tool	Description
get_notifications	Returns recent notifications, optionally filtered by read state, level, time, or cursor.
get_job_output_by_id	Returns pipeline or job-stream output when the caller is allowed to read that job.
search_ant_design_controls	Searches available Ant Design and PSBlazor portal controls and returns their parameters.

Tool limits and run targets

PowerShell Universal supports optional per-minute MCP tool rate limiting and an optional concurrent-job limit for AI tool jobs. Configure these limits in the server settings to protect the instance from excessive tool activity.

Custom AI Tools can also define Run In and Run On targets. These settings control the environment and computer or computer group used when a custom tool starts a job. They do not change the authorization requirements of the built-in MCP tools listed on this page.

Examples

Use resource discovery before editing configuration:

list_resource_types
list_resources(type: "scripts", filter: "Inventory*")
get_file_content(path: "scripts/Inventory.ps1")

After an authorized update, reload and inspect notifications:

set_file_content(path: "scripts/Inventory.ps1", content: "<new script content>")
reload_resource(resourceType: "scripts")
get_notifications(unreadOnly: true, levels: "Warning,Error")