Automating the Software Development Lifecycle with LLM Coding Agents: A Webhook-Based Approach

• Arthur Claro UCB

DOI: https://doi.org/10.5753/eniac.2026.30089

Resumo

This paper reports on the design and deployment of a system that integrates an autonomous Large Language Model (LLM) coding agent with a corporate issue tracker. Moving a ticket to a dedicated status, or mentioning the agent in a comment, triggers a webhook that hands the task to the agent. The agent reads the ticket, creates a branch, implements the change, runs lint, build and the test suite, opens a pull request (PR), and updates the ticket. We describe the event-driven architecture (webhook server, secure edge routing, process supervision and per-card concurrency control) and its incremental deployment in a real software organization. Across two detailed observation windows we reconstructed several dozen runs; seven exercised the full status-triggered flow. A third window is reported in aggregate only. Two non-test tickets reached merge, one with no human code rework and one after two rounds of review; the remaining status runs were declined test PRs, a still-open pull request, a mid-run API-quota abort, and one killed by the 30-minute watchdog after a concurrency conflict between the two triggers. The agent opened PRs in about fifteen minutes, while end-to-end time to merge was measured in days and dominated by human review, so this reflects latency to review rather than total cost. The twelve runs we could price in the second window cost US$ 23.26 in total. A recurring lesson is that most of the engineering effort went into operational integration, such as the network edge, TLS termination and event delivery, rather than into the agent’s reasoning.

Referências

Alshahwan, N. et al. (2024). Automated unit test improvement using large language models at Meta. In Companion Proc. FSE, Industry track.

Becker, J., Rush, N., Barnes, B., and Rein, D. (2025). Measuring the impact of early-2025 AI on experienced open-source developer productivity. arXiv preprint arXiv:2507.09089.

Chen, M. et al. (2021). Evaluating large language models trained on code. arXiv preprint arXiv:2107.03374.

Cognition AI (2024). Introducing Devin, the first AI software engineer. Available at [link]. Accessed 6 June 2026.

Frömmgen, A. et al. (2024). Resolving code review comments with machine learning. In Proc. ICSE-SEIP.

GitHub (2025). GitHub Copilot: meet the new coding agent. Available at [link]. Accessed 6 June 2026.

Hohpe, G. and Woolf, B. (2003). Enterprise Integration Patterns: Designing, Building, and Deploying Messaging Solutions. Addison-Wesley, Boston.

Hong, S. et al. (2024). MetaGPT: meta programming for a multi-agent collaborative framework. In Proc. ICLR.

Jimenez, C. E., Yang, J., Wettig, A., Yao, S., Pei, K., Press, O., and Narasimhan, K. (2024). SWE-bench: can language models resolve real-world GitHub issues? In Proc. ICLR.

Liu, J., Wang, K., Chen, Y., Peng, X., Chen, Z., Zhang, L., and Lou, Y. (2024). Large language model-based agents for software engineering: a survey. To appear in ACM TOSEM. arXiv:2409.02977.

Peng, S., Kalliamvakou, E., Cihon, P., and Demirer, M. (2023). The impact of AI on developer productivity: evidence from GitHub Copilot. arXiv preprint arXiv:2302.06590.

Qian, C. et al. (2024). ChatDev: communicative agents for software development. In Proc. ACL.

Storey, M.-A. and Zagalsky, A. (2016). Disrupting developer productivity one bot at a time. In Proc. FSE, Visions and Reflections track.

Wang, X. et al. (2025). OpenHands: an open platform for AI software developers as generalist agents. In Proc. ICLR. arXiv:2407.16741.

Yang, J., Jimenez, C. E., Wettig, A., Lieret, K., Yao, S., Narasimhan, K., and Press, O. (2024). SWE-agent: agent-computer interfaces enable automated software engineering. In Proc. NeurIPS.

Yao, S., Zhao, J., Yu, D., Du, N., Shafran, I., Narasimhan, K., and Cao, Y. (2023). ReAct: synergizing reasoning and acting in language models. In Proc. ICLR.

Zhang, Y., Ruan, H., Fan, Z., and Roychoudhury, A. (2024). AutoCodeRover: autonomous program improvement. In Proc. ISSTA.