From Jira ticket to feature environment: how we built a multi-agent AI developer

说明与用途

Codey 是一个多代理 AI 系统，自动完成从 Jira 任务到功能部署的完整开发周期，成功率 75%，平均交付时间约 30 分钟。

问题

Integrating AI coding agents into a real development process to automate end‑to‑end feature delivery while maintaining quality and developer trust.

结果

Codey automates 7% of tasks, reduces time from coding start to feature environment to ~30 minutes, and achieves a 75% successful completion rate across 250 tasks.

工作流

• 1. Task enters Jira and is labeled for Codey or marked AI skip.
• 2. Codey identifies related GitLab repositories and project settings.
• 3. Six AI agents (Researcher, Planner, Developer, Reviewer, QA, Verifier) execute the task in sequence.
• 4. Codey runs automated checks (tests, linters, security, AI review).
• 5. If checks fail, task returns to Developer for rework; max three iterations.
• 6. Upon passing, Codey creates a merge request and deploys a feature environment.
• 7. Requester reviews in the preview, confirms or rejects; final release proceeds with human approval.

证据

250 tasks have gone through the system

来源

75% of them were completed successfully

来源

~30 minutes is the typical time from the start of coding to a deployed feature environment

来源

Codey currently works well with Node.js, React and Python projects

来源