kestra_io

A new engineering role is showing up on job boards in 2026. Here is what a frontier engineer actually does, how it differs from an ML or platform engineer.

Kestra's cloud catalog used to mean three providers. In 2026 it stopped meaning that. Where every cloud plugin stands today and what is still missing.

What we fixed and hardened in the 2.0 cycle: authentication, information disclosure, CSRF and XSS, SSRF filtering, upload protection, secrets masking, and why workers no longer touch the database.

Kestra 2.0 turns the plugin system from an integration catalogue into an extension point for the platform itself: policies, file previews, log storage, the queue, and now the UI.

Cases in Kestra 2.0 turn repeated failures into one owned incident, with an SLA, the evidence attached, and the fix as a button. What it gives you, and how to run it in production.

For data engineers evaluating Airflow or Dagster alternatives: how Kestra 2.0's architecture removes the worker-to-database dependency, and where it still falls short.

One control plane across data, infrastructure, applications and business processes. Why Kestra 2.0 needed a new engine, what it unlocks, and what changes when you upgrade.

Flow configuration, incidents and promotion between environments can now be governed with Policies, Cases and Promote. Flows can be saved as Drafts and published when ready, and the AI Copilot is now a persistent chat panel with memory.

Kestra 2.0 doubles sustained throughput to 4000 executions per minute on the same Postgres, with lower latency and a flat p99. Here is what changed in the engine.

Kestra 2.0 does not ship you an agent. One trigger turns any flow into an MCP tool your existing agents can call, and every agent call becomes a normal execution with the same logs, RBAC, and audit trail.