Kiro Product
CLI
Feature Description
Add support for "routines" — saved Kiro configurations (prompt + repos + tools/skills) that run automatically based on triggers, similar to Claude Code's Routines feature.
This could be delivered incrementally:
Phase 1 — Local scheduled tasks (cron)
The simplest starting point. Let users define a named task with a prompt, working directory, and a cron schedule. The task runs locally while Kiro is open, with full access to the local filesystem, MCP servers, skills, and tools. Claude Code supports this today via Desktop scheduled tasks (hourly, daily, weekdays, weekly, custom cron) and via /loop for session-scoped polling. Even just this level of support would be a big win for Kiro users.
Phase 2 — Cloud routines with richer triggers
Once local scheduling exists, the next step would be cloud-hosted routines that run on managed infrastructure (no laptop required) with additional trigger types:
- Cron schedule — recurring cadence (hourly, nightly, weekly, custom cron) or one-off at a specific future time
- API/webhook — a dedicated HTTP endpoint per routine that external systems can POST to (CI/CD pipelines, monitoring alerts, deploy scripts, internal tools), with optional runtime context passed in the request body
- GitHub events — react automatically to repository events like PR opened, PR merged, release published, with filters (by author, label, base branch, draft status, etc.)
Key capabilities for the full vision:
- Define once, run many times — save a prompt, target repos, skills, MCP servers, and environment config as a named routine
- Multiple triggers per routine — combine schedule + API + GitHub triggers on the same routine
- Run context — each trigger can pass runtime context (e.g., an alert body, a PR URL) alongside the saved prompt
- Full Kiro capabilities — routines have access to skills, MCP servers, and tools just like an interactive session
- Connectors — integrate with external services (Slack, Jira, etc.) so routines can read from and write to external systems
- Environment configuration — control network access, environment variables, and setup scripts per routine
- Branch safety — by default, only push to prefixed branches (e.g.,
kiro/), with an opt-in for unrestricted branch pushes
- Manageable — list, edit, pause/resume, delete routines; view run history and interact with completed runs
Use Case
-
Automated code review — GitHub trigger on pull_request.opened runs the team's review checklist, leaves inline comments, and posts a summary comment.
-
Backlog maintenance — Nightly schedule reads new Jira issues, applies labels, assigns owners based on code area, and posts a summary to Slack.
-
Deploy verification — CD pipeline calls the routine's API endpoint after each deploy. Routine runs smoke checks, scans error logs, and posts go/no-go to the release channel.
-
Alert triage — Monitoring tool POSTs to the routine's webhook when an error threshold is crossed. Routine correlates the stack trace with recent commits and opens a draft PR with a proposed fix.
-
Docs drift detection — Weekly schedule scans merged PRs, flags documentation that references changed APIs, and opens update PRs.
-
Local dev loop — Even without cloud infrastructure, a local cron task that runs tests every 30 minutes, checks for lint regressions, or polls a staging environment would be immediately useful.
Additional Context
Claude Code has a tiered approach to this problem worth studying:
/loop — session-scoped cron scheduling in the CLI (expires after 7 days, lost on new session)- Desktop scheduled tasks — local cron-only scheduling that persists across sessions (requires app open + machine awake)
- Cloud Routines — full trigger support (cron + API + GitHub events) on managed infrastructure
Full docs: https://docs.anthropic.com/en/docs/claude-code/routines
Starting with local cron-based scheduled tasks would deliver immediate value without requiring cloud infrastructure. The richer trigger types (API, GitHub events) and cloud execution can follow as the platform matures.
Kiro Product
CLI
Feature Description
Add support for "routines" — saved Kiro configurations (prompt + repos + tools/skills) that run automatically based on triggers, similar to Claude Code's Routines feature.
This could be delivered incrementally:
Phase 1 — Local scheduled tasks (cron)
The simplest starting point. Let users define a named task with a prompt, working directory, and a cron schedule. The task runs locally while Kiro is open, with full access to the local filesystem, MCP servers, skills, and tools. Claude Code supports this today via Desktop scheduled tasks (hourly, daily, weekdays, weekly, custom cron) and via
/loopfor session-scoped polling. Even just this level of support would be a big win for Kiro users.Phase 2 — Cloud routines with richer triggers
Once local scheduling exists, the next step would be cloud-hosted routines that run on managed infrastructure (no laptop required) with additional trigger types:
Key capabilities for the full vision:
kiro/), with an opt-in for unrestricted branch pushesUse Case
Automated code review — GitHub trigger on
pull_request.openedruns the team's review checklist, leaves inline comments, and posts a summary comment.Backlog maintenance — Nightly schedule reads new Jira issues, applies labels, assigns owners based on code area, and posts a summary to Slack.
Deploy verification — CD pipeline calls the routine's API endpoint after each deploy. Routine runs smoke checks, scans error logs, and posts go/no-go to the release channel.
Alert triage — Monitoring tool POSTs to the routine's webhook when an error threshold is crossed. Routine correlates the stack trace with recent commits and opens a draft PR with a proposed fix.
Docs drift detection — Weekly schedule scans merged PRs, flags documentation that references changed APIs, and opens update PRs.
Local dev loop — Even without cloud infrastructure, a local cron task that runs tests every 30 minutes, checks for lint regressions, or polls a staging environment would be immediately useful.
Additional Context
Claude Code has a tiered approach to this problem worth studying:
/loop— session-scoped cron scheduling in the CLI (expires after 7 days, lost on new session)Full docs: https://docs.anthropic.com/en/docs/claude-code/routines
Starting with local cron-based scheduled tasks would deliver immediate value without requiring cloud infrastructure. The richer trigger types (API, GitHub events) and cloud execution can follow as the platform matures.