Solo SaaS dev automates customer support with Claude Code Schedule
Developer kanta13jp1 shares how Claude Code CLI's Schedule feature can run an hourly `cs-check` task that auto-replies to FAQ tickets, patches simple bugs and commits fixes, escalates complex cases, and pings live endpoints — all without human intervention.
Score breakdown
In a Dev.to post, solo SaaS developer kanta13jp1 outlines a system that uses Claude Code CLI's Schedule feature to handle customer support automatically. A `cs-check` task runs on a `cron: "0 * * * *"` schedule, fetches unreplied tickets from Supabase, and routes them into three paths: FAQ auto-reply (similarity score > 0.7), bug investigation and fix, or human escalation. When a fixable bug is detected, Claude patches the code, runs `flutter analyze`, commits with a descriptive message, and replies to the user. The author claims the system cuts average response time from 24 hours to under 1 hour and reduces human intervention to roughly 20% of tickets.
- 01Claude Code CLI's Schedule feature runs a `cs-check` task every hour using a `cron: "0 * * * *"` expression defined in `.claude/schedule.yaml`.
- 02Tickets are fetched from Supabase and routed into three cases: FAQ auto-reply (similarity > 0.7), bug investigation/fix, or escalation.
- 03When a bug is auto-fixed, Claude runs `flutter analyze`, commits with `git add -p`, pushes to `main`, and replies to the user automatically.
Writing on Dev.to, solo SaaS developer kanta13jp1 describes an automated customer support pipeline built around Claude Code CLI's Schedule feature. A `.claude/schedule.yaml` config defines a `cs-check` task on a `cron: "0 * * * *"` schedule, which every hour queries Supabase for open, unreplied support tickets and routes each one through a three-branch decision tree: tickets matching a FAQ entry at a similarity score above `0.7` receive an automatic reply; tickets containing bug-signal keywords like `'error'`, `'broken'`, or `"can't"` trigger a code investigation; and billing, refund, or otherwise complex tickets are escalated and flagged for human review.
The bug-fix branch is the most technically ambitious part of the workflow.
The bug-fix branch is the most technically ambitious part of the workflow. When Claude identifies a simple, fixable issue, it applies a patch, runs `flutter analyze lib/` to confirm zero errors, stages changes with `git add -p`, commits with a structured message (e.g., `"fix: null check in home dashboard loader"`), pushes to `main`, and sends the user a reply noting that a deploy is in progress. Every session also pings a list of live endpoints with a 10-second `AbortSignal.timeout` and logs any non-OK responses as infrastructure alerts — providing lightweight uptime monitoring with no additional tooling. All activity is written to timestamped markdown logs at `docs/cs-notes/YYYY-MM-DD-HH.md`.
The author's self-reported metrics frame the ROI clearly: average response time drops from 24 hours to under 1 hour, human intervention falls from 100% to roughly 20% of tickets, and bug-to-fix latency shrinks from days to minutes. For solo builders who treat every support ticket as lost build time, this pattern positions scheduled agentic tasks as one of the highest-leverage automation investments available.
Key facts
- 01Claude Code CLI's Schedule feature runs a `cs-check` task every hour using a `cron: "0 * * * *"` expression defined in `.claude/schedule.yaml`.
- 02Tickets are fetched from Supabase and routed into three cases: FAQ auto-reply (similarity > 0.7), bug investigation/fix, or escalation.
- 03When a bug is auto-fixed, Claude runs `flutter analyze`, commits with `git add -p`, pushes to `main`, and replies to the user automatically.
- 04A per-session markdown log is written to `docs/cs-notes/YYYY-MM-DD-HH.md` covering replied, fixed, and escalated tickets.
- 05Live endpoint health checks use `AbortSignal.timeout(10000)` and append infrastructure alerts to the same log — no extra monitoring tools required.