I built a post-deploy security agent with Kestra that scans package dependencies, monitors live browser network requests using Playwright, and rolls back Vercel deployments automatically when something looks unsafe. The biggest thing Kestra helped with was removing all the boilerplate. Scheduling, webhook validation, running Docker containers as tasks, passing values between flows using KV Store, Slack notifications, HTTP calls to Gemini and Vercel APIs, all of that was just a task type and a few YAML lines. None of it needed custom code. The actual work stayed focused on the scan logic and rollback decision, not the infrastructure around it. The execution visibility also made a real difference. When something broke, I could see exactly which task failed, what it returned, and where the flow stopped. That made debugging much faster than digging through terminal logs.

I really like the workflows in Kestra, especially how they enhance documentation. In traditional DevOps setups, setting up cron jobs can be cumbersome, but Kestra simplifies this process. I enjoy the decorative YAML-based flow definitions which make workflows self-documenting. Features like conditional branches and test output functions allow for passing data between tasks and configuring inputs, which provides meaningful execution logic. The UI timeline is incredibly easy to use, allowing me to execute and diagnose failures without needing to dig into the logs across multiple platforms.

I used Kestra while building AI workflow projects like API Roulette, where multiple APIs, AI-generated outputs, and asynchronous tasks needed to coordinate reliably. It helped me move away from messy backend scripts into structured workflows with retries, execution tracking, logging, and DAG-based visibility. I especially liked the workflow observability features, since watching executions through DAGs, logs, retries, and task outputs made debugging much easier compared to traditional scripts. The Docker setup and onboarding experience felt very smooth, and integrating Kestra with REST APIs and AI workflows was straightforward. Performance during local development was also surprisingly good even with larger multi-step workflows. Since retries, scheduling, logging, and monitoring are already built in, it reduced the need for extra backend tooling and made the overall development experience much cleaner. For the features and workflow visibility Kestra provides out of the box, I feel it delivers a lot of value, especially for automation and AI workflow projects that would otherwise require much more custom backend setup.