Edit

AL Language Model tools for Visual Studio Code

Applies to: AL Language extension 17.0 and later

The AL Language extension registers a set of Language Model Tools with the Visual Studio Code Language Model Tools API. These tools are available to GitHub Copilot in Agent mode and to any Copilot-compatible extension running inside Visual Studio Code. They allow an AI agent to perform core AL development tasks—building, publishing, downloading symbols, searching symbols, reading diagnostics, and debugging—directly from a conversation.

Available tools

Tool Description
al_build Build the AL project and generate the .app package.
al_publish Publish the extension to Business Central.
al_downloadsymbols Download dependent symbol packages from the configured BC server or global sources.
al_symbolsearch Search AL symbols across the project and its dependencies.
al_getdiagnostics Retrieve compilation diagnostics from the Problems panel.
al_debug Start a debug session without republishing.
al_setbreakpoint Add, remove, or toggle a breakpoint at a specific file and line.
al_snapshotdebugging Manage snapshot debugging sessions.

Using tools in Copilot Chat

When GitHub Copilot is in Agent mode, it can invoke AL tools automatically in response to natural-language requests. You can also reference a specific tool explicitly by prefixing its name with # in the chat input:

#al_build
#al_symbolsearch
#al_getdiagnostics

Example prompts

Your prompt What happens
"Build my AL project." Copilot calls al_build with scope="current".
"Build everything in the workspace." Copilot calls al_build with scope="all".
"Publish my extension to my sandbox." Copilot calls al_publish using connection details from launch.json.
"Download symbols." Copilot calls al_downloadsymbols.
"Find all codeunits related to posting." Copilot calls al_symbolsearch with query="posting" and kinds=["Codeunit"].
"Show me all errors in the project." Copilot calls al_getdiagnostics with scope="project" and severities=["error"].
"Start debugging without republishing." Copilot calls al_debug.
"Set a breakpoint on line 42 of CustomerMgt.Codeunit.al." Copilot calls al_setbreakpoint with lineNumber=42 and action="add".
"Start snapshot debugging." Copilot calls al_snapshotdebugging with action="initialize".

Tool chaining

Copilot can chain AL tools automatically in a single conversation turn. For example:

  • If al_build reports errors, Copilot might call al_getdiagnostics to retrieve the full list, suggest code fixes, rebuild, and then publish—all without further prompting.
  • If you ask Copilot to investigate a bug, it might call al_symbolsearch to locate the relevant code, set a breakpoint with al_setbreakpoint, and then start debugging with al_debug.

This behavior is guided by workflow chains that the AL Language Server defines for each tool. The chains describe likely next steps after success or failure, enabling Copilot to make coherent, multi-step decisions.

Connection settings

The Visual Studio Code tools read connection settings from the active launch.json configuration in your workspace. You don't need to specify server URLs, tenant IDs, or authentication parameters in the chat—Copilot derives them from your workspace configuration.

To target a different environment, update launch.json or switch the active launch configuration in Visual Studio Code before invoking the tool.

Downloading symbols from global sources

The al_downloadsymbols tool supports a globalSourcesOnly option that downloads Microsoft platform and base application symbols from Microsoft NuGet feeds and AppSource without requiring a connection to a live Business Central server. This is useful when:

  • You're working offline.
  • You want to build without configuring a server connection.
  • You're running in a CI environment where a BC server isn't available.

To use this in Copilot Chat: "Download symbols from global sources only."

Prerequisites

  • The AL Language extension version 17.0 or later must be installed in Visual Studio Code.
  • GitHub Copilot must be installed and signed in.
  • Your workspace must contain a valid AL project with an app.json file.
  • A launch.json configuration is required for tools that connect to a Business Central server (al_publish, al_downloadsymbols without globalSourcesOnly).

AI agent tools overview
AL MCP Server reference
al_build
al_publish
al_downloadsymbols
al_symbolsearch
al_getdiagnostics
al_debug, al_setbreakpoint, al_snapshotdebugging

  • Last updated on