EUVIMEDEUVIMED entdecken
Open Access · CC BY-SA 4.0

Comparative Evaluation of Agentic Workflow Capabilities in AI IDE Agents for Web-Based Learning Media Development

Erlangga Aditia, Ulva Elviani

Journal of Applied Informatics and Computing · 12.06.2026 · 10 (3) · 2556-2567

DOI: 10.30871/jaic.v10i3.12841 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Applied Informatics and Computing, online erschienen am 12.06.2026. Verfasst von Erlangga Aditia und 1 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref). Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The rapid development of agentic IDEs calls for evaluation approaches that assess not only final outputs but also the agentic workflow enacted during software development. This study comparatively evaluates the workflow capabilities of four AI IDE agents, namely Cursor, Windsurf, Trae, and Antigravity, within the five-stage benchmark of developing a web-based learning media application, Next-Gen SPLDV. A descriptive comparative evaluation was conducted using five agentic maturity metrics: task decomposition (DC), tool-use effectiveness (TSR), autonomous recovery capability (ARC), human intervention cost (HIC), and time completion efficiency (TCT), complemented by interaction logs and internal artifacts. The findings indicate distinct performance trade-off profiles across systems. Antigravity appeared relatively more stable descriptively (TSR 96.0%; HIC 3; ARC 8), whereas the other systems exhibited context-dependent strengths: Cursor showed more selective tool use, Trae was efficient in several stages but more vulnerable during database integration, and Windsurf was more exploratory but required higher intervention and recovery effort. Qualitative evidence further suggests that these differences were associated with variations in plan-execute-verify strategies and error-response behavior. Overall, the evaluation of AI IDE agents is better interpreted as a contextual map of workflow trade-offs rather than the identification of a single winner across all settings.

Zitieren

APA

Aditia, E., & Elviani, U. (2026). Comparative Evaluation of Agentic Workflow Capabilities in AI IDE Agents for Web-Based Learning Media Development. Journal of Applied Informatics and Computing, 10(3), 2556-2567. https://doi.org/10.30871/jaic.v10i3.12841

Vancouver

Aditia E, Elviani U. Comparative Evaluation of Agentic Workflow Capabilities in AI IDE Agents for Web-Based Learning Media Development. Journal of Applied Informatics and Computing. 2026;10(3):2556-2567. doi:10.30871/jaic.v10i3.12841

BibTeX
@article{Aditia2026,
  author = {Aditia, Erlangga and Elviani, Ulva},
  title = {Comparative Evaluation of Agentic Workflow Capabilities in AI IDE Agents for Web-Based Learning Media Development},
  journal = {Journal of Applied Informatics and Computing},
  year = {2026},
  volume = {10},
  number = {3},
  pages = {2556--2567},
  doi = {10.30871/jaic.v10i3.12841}
}
Kostenlos starten →