Kiro introduced workflows for multi-agent plan, code and review pipelines

Archive entry: this covers an event from September 30, 2026, and we published it on November 8, 2026. The details come from sources dated at the time and may have changed since.

On September 30, 2026, Kiro introduced workflows in a Kiro blog post introducing workflows, written by Romain Dura and Doug Clauson. A workflow was a graph of agent steps, loops, parallel branches and sequences that carried out a complex task with multiple agents.

What workflows added

• Each step got its own session and clean context, so a reviewer judged output without seeing the coder's reasoning.
• Workflows ran asynchronously, and step sessions could be paused, resumed or redirected mid-run.
• Reusable recipes were JSON (YAML also worked), built from step, parallel and repeat constructs, and Kiro could generate them.
• Workflows were available in the IDE, CLI and Web on one runtime, and cloud sessions could run them.
• Built-in recipes included investigate, a read-only probe, and publish-pr, which opened a PR, followed it to merge and retried failed CI checks.

What it meant for people who build with AI

Kiro users could automate plan-implement-review loops with separate agents instead of one long chat. Because each step started from a clean context, a review step could check the code without inheriting the reasoning that produced it, and Kiro could generate the recipes.

Availability and cost

Workflows launched as opt-in, enabled with the setting kiroAgent.workflows.enabled. They used the same credit model as other Kiro agent work, so complex runs could consume more credits.

Sources

• Introducing Kiro workflows (Kiro, September 30, 2026, official)

Written with AI assistance, strictly from the sources listed above. Before it went live, automated checks verified its dates, sources and every language edition.