Effectiveness of AL-Specific Tooling

Determine whether providing coding agents with AL-specific tooling, including AL MCP and AL LSP capabilities such as compilation, symbol search, and go-to definition, yields meaningful gains in Business Central/AL task performance.

Background

In the default BC-Bench setting, agents interact with the AL codebase as plain text and cannot directly access the underlying Business Central development environment. Selected experiments expose portions of that environment through AL MCP, while the failure analysis suggests that repository navigation, context retrieval, compilation, and test feedback may affect performance. The paper therefore leaves unresolved whether controlled interventions using AL MCP and AL LSP produce meaningful improvements.

References

Whether these interventions yield meaningful gains is an empirical question that BC-Bench enables answering.

BC-Bench: Evaluating Agentic Engineering in a Domain-Specific Language for ERP  (2608.20851 - Sun et al., 21 Aug 2026) in Section 6, Future Work, subsection “Experiments”