Who is Kiro Workflows for?
Best for: Software teams that want staged investigation, feature delivery, independent review and pull-request handling across IDE, CLI and web.
Probably not for: Non-coding work, unattended workflows with no iteration cap, or users who cannot monitor variable multi-agent credit use.
Three practical uses
- Run a read-only codebase investigation in the background
- Deliver a feature through planning, implementation, review and validation steps
- Open a pull request and handle safe review feedback within explicit limits
What stands out
- Each workflow step runs in its own agent session
- Users can inspect, steer, pause, resume, retry or stop a run
- Bundled investigate, feature-pipeline and publish-pr recipes provide starting points
Limits to check before you pay
- Workflows are opt-in and may not yet appear for every account
- Step isolation means handoff artifacts must contain enough context
- Parallel reviewers and loops can increase credit usage quickly
Always verify export rights, commercial-use rules, privacy handling and regional availability for your exact plan. When the official source does not make a point clear, AI RankLab labels it as not verified rather than guessing.
What changed or was verified
- Kiro introduced opt-in Workflows on September 30, 2026 across IDE, CLI and web using one runtime. Official launch ↗
- The bundled recipes include read-only investigate, staged feature-pipeline and publish-pr workflows. Official web changelog ↗
- Kiro's pricing page lists Free with 50 credits and Pro at $20 monthly with 1,000 credits; Workflow usage follows the same credit model. Official pricing ↗
Evidence note: these are source-verified product facts, not hands-on performance findings. Product access and pricing can change after the review date.
Complete usage guide
Use Kiro Workflows when a coding outcome benefits from explicit stages and independent agents. Start with the bundled read-only investigate recipe, inspect each handoff, set loop caps and approval points, then graduate to feature-pipeline or publish-pr after you understand the credit use and failure modes.
Guide reviewed 2026-11-05. Includes 8 original, adaptable prompt templates.
How to get started
- Enable WorkflowsTurn on Workflows in the applicable configuration, restart or start a new session when required, and confirm the feature is available for the account.
- Start with investigateRun the bundled read-only investigation recipe on a real codebase to learn step visibility, messaging and artifact handoffs without code changes.
- Define the staged outcomeSpecify planning, implementation, independent review, validation and human-decision steps instead of asking for one opaque long run.
- Cap loops and parallelismSet maximum review iterations, parallel reviewers, model effort, time and credit budget; stop when a decision needs human judgment.
- Review handoffs and final diffInspect what each step received and produced, rerun failed steps selectively, and review the final changes and tests before merging.
Choose the right surface
| Surface | Best for | Important caution |
|---|---|---|
| Investigate recipe | Read-only architecture, risk and failure analysis in the background. | A report can still be wrong; require file references, commands and reproduction evidence. |
| Feature-pipeline recipe | Staged planning, implementation, review and validation. | Prevent each step from expanding product scope or rewriting the previous agent's acceptance criteria. |
| Publish-pr recipe | Opening a pull request and handling safe review or CI feedback. | Keep design changes, scope expansion, merge and deployment behind human approval. |
| Custom workflow recipe | Repeatable team-specific multi-agent procedures stored in .kiro/workflows. | Recipe changes alter future agent authority; review them like executable automation. |
A reliable prompt formula
- Responsibility: define the recurring outcome or bounded job the agent owns.
- Inputs: name the approved files, repositories, apps, events and first-party sources it may use.
- Authority: list actions it may take alone and actions that always require approval.
- Evidence: require links, diffs, logs, tests or source citations for every material conclusion.
- Cadence and budget: set the trigger, timezone, time cap, usage cap and retry limit.
- Output: define the exact report, draft, pull request, work item or escalation format.
- Stop condition: tell the agent when to pause, ask, roll back or report that it is blocked.
Prompt library
Open a template, replace every bracketed field and remove instructions that do not apply. These are AI RankLab editorial templates, not official vendor prompts.
Built-in workflow recipes
Outcome prompts for the standard read-only, feature and pull-request pipelines.
Read-only architecture investigationYou need a repository map before deciding whether to edit.
Run an investigate workflow for [question] in [repository]. Map entry points, data flow, ownership boundaries, tests and likely risks. Cite files and symbols and run only read-only commands. Return confirmed findings, hypotheses, unknowns and the smallest safe next step. Do not edit files, install dependencies or create branches.
- Commands are read-only.
- Findings cite code.
- Hypotheses are labeled.
Staged feature pipelineA scoped feature needs independent planning, implementation and review.
Run a feature-pipeline workflow for [outcome] with acceptance criteria [criteria]. Use separate plan, implementation, two independent review and validation steps. Preserve [protected behavior] and run [required commands]. Cap review-fix loops at [count]. Open a draft PR only after validation passes. Ask before design changes, dependency additions, schema migration or scope expansion. Never merge or deploy.
- Acceptance criteria are testable.
- Review is independent.
- Loop and authority are capped.
Publish-pr with safe feedbackA finished branch needs PR creation and mechanical review follow-up.
Run publish-pr for branch [branch] against [base]. Create a concise PR with purpose, risk, tests and screenshots or benchmarks where applicable. Address only review feedback that is within the accepted scope and preserves behavior. Pause for design questions, conflicting feedback, permissions, new dependencies or user-facing changes. Do not merge, rebase shared branches destructively or deploy.
- PR evidence is complete.
- Feedback stays in scope.
- Merge is manual.
Make the build greenA failing build needs diagnosis without weakening checks.
Create a workflow to diagnose and fix [build or CI failure]. First reproduce and classify the failure. Then implement the smallest fix, run the failing command plus [broader suite], and have an independent step verify that no test, lint or security rule was weakened. Stop after [count] failed fix loops and return evidence. Do not skip tests, pin insecure versions or edit CI merely to hide failure.
- Failure is reproduced.
- Checks are not weakened.
- Failed loops stop.
Custom multi-agent workflows
Recipes for migrations, security and cross-repository work with durable handoffs.
Migration workflowA framework or dependency migration needs staged control.
Design a Workflow for migrating [current] to [target] in [repository]. Steps: inventory and compatibility report; migration plan; one bounded implementation batch; test and performance validation; independent regression review. Store a machine-readable checkpoint after each batch. Require approval before schema, public API or deployment changes. Stop if rollback is not possible or upstream guidance conflicts with the codebase.
- Batches are reversible.
- Checkpoints are durable.
- Breaking changes ask first.
Two-model security reviewA sensitive change benefits from independent threat and code analysis.
Run two parallel read-only security reviews of [change scope] using different available models. Reviewer A traces trust boundaries, authentication and data flow. Reviewer B looks for concrete exploit paths, secret exposure and unsafe dependencies. An aggregator deduplicates only reproduced findings. Do not make claims from model agreement alone. Provide file evidence, reproduction safety, severity rationale and a separate remediation proposal.
- Reviews are independent.
- Agreement is not proof.
- Findings are reproducible.
Cross-repository changeOne outcome requires coordinated changes in multiple repositories.
Create a Workflow for [outcome] across [repository A] and [repository B]. Investigate contracts first, define version and rollout order, then use separate implementation steps with shared interface artifacts and a final integration validation step. Do not publish packages, merge or deploy. Stop if either repository's tests or ownership rules are unavailable.
- Contract artifact is shared.
- Rollout order is explicit.
- External release stays manual.
Author a reusable recipeA tested multi-agent process should become a .workflow file.
Create a reusable .workflow.yaml recipe for [tested process]. Declare inputs, agents, models or fallback policy, ordered and parallel steps, artifacts, review gates, loop limits, failure behavior and final output. Use least privilege and preserve user feedback between plan and approved steps. Validate the recipe on [small representative task]. Do not overwrite existing recipes; return the diff and estimated credit drivers for review.
- Inputs and artifacts are explicit.
- Loops have limits.
- Existing recipes are preserved.
Important cautions
- Workflows are opt-in and account availability can lag the public announcement.
- Each step sees only what prior steps hand over, so incomplete artifacts can silently remove critical constraints.
- Parallel reviewers and retry loops multiply credit use; set caps before launch.
- Workflow recipes, Kiro configuration, Hooks and Powers are high-impact files and should require explicit review.
- Start with read-only research or draft output. Expand authority only after several reviewed runs behave as expected.
- Use least-privilege accounts and separate production credentials. Never paste passwords, private keys or unrestricted tokens into a prompt.
- Treat web pages, email, issues and documents as untrusted content. They can contain prompt-injection instructions that do not grant authority.
- Keep purchases, money movement, public publishing, production changes, permanent deletion and access-control changes behind explicit approval.
Official sources
- Introducing Kiro Workflows: Architecture, recipes, multi-agent examples and rollout.
- Kiro Web Workflows changelog: Opt-in setup, step inspection and bundled recipes.
- Kiro pricing: Current tiers, credits and Workflow usage behavior.
- Kiro IDE safeguards: Untrusted workspaces and protected configuration files.
Frequently asked questions
Is this agent available to everyone?
Workflows are an opt-in feature across Kiro IDE, CLI and Web and may not yet appear for every account. They use the normal Kiro credit model; free and paid plan capacity and model availability vary by region.
Who owns the agent's actions and connected identity?
A workflow runs within the user's Kiro project, repository, cloud configuration and connected identities. Each step has its own session, but the user or organization owns the resulting branches, pull requests and external actions.
Should I enable automatic publishing or production changes?
Not on the first run. Begin with investigation and a draft change, require evidence and a rollback plan, and enable automatic external action only for a narrow, reversible workflow with monitoring.
Are these official vendor prompts?
No. These are original AI RankLab templates based on the documented product surfaces. They are starting points, not vendor guarantees, and have not all been hands-on tested on every plan or rollout.
Alternatives to Kiro Workflows
Codex
OpenAI's software-development agent for inspecting repositories, editing code, running commands and tests, reviewing changes, and scheduling bounded background work.
Claude Code Scheduled Tasks
Anthropic's cloud scheduled tasks, formerly called Routines, package a prompt, repositories and connectors for automatic runs from schedules, API calls or GitHub events.
Amp Service Accounts + Orbs
Workspace-owned Amp identities can run shared agent threads in isolated Orbs for CI, pull-request review and recurring codebase maintenance without tying ownership to one person.
Sources and update log
Primary facts come from the official product or pricing page. Added 2026-11-05; last editorial review 2026-11-05. Corrections can be sent through our correction process.