How one freelancer billed 3 client apps on autopilot — and what their "ops dashboard" looks like.

The portfolio

Three products, all on retainer:

• Client A — fitness studio booking SaaS, ~600 paying users, MRR ~$12k. Built by me, running 11 months.
• Client B — accounting tool for small e-commerce shops, ~200 users, MRR ~$8k. 6 months in production.
• Client C — coaching marketplace, ~150 active coaches, MRR ~$8k. 4 months in production.

I charge $2k/month flat per client for ops + small features. I do not bill hours. We agreed on a monthly cap of 12 hours per client; I almost never use them.

The morning routine (15 minutes)

Every weekday at 8 AM, the agent posts a summary to a private Slack channel for each client. Same template across all three:

📊 [Client A] Morning report — Mon Apr 27

Yesterday: 87 sign-ups (+12% vs avg), 4 churned
MRR: $11,890 (+$140)
Errors: 2 (both 502s on /export, retried successfully)
Slow queries: 1 (find_bookings, 1.2s p95 — flagged)
Latest deploy: 2 days ago, no regressions
Pending tasks from you (3): see this thread

I read three of those over coffee. Most days, that's it. The "pending tasks" line is the agent reminding me of things I said I'd do, not things it's blocked on.

What only the agent does

Things I used to do every week, that the agent now does:

• Database backups verified. Nightly D1 export to R2. Agent restores it to a sandbox once a week and runs a smoke test.
• Failed payment retries. Stripe webhook → agent retries 3, 7, 14 days, drops to grace period if all fail.
• Onboarding emails. Cron-triggered, sent through Resend, content evolves with feature releases.
• Weekly client digest email. Different from my morning Slack — this goes to the client's CEO with a polished summary.
• Cleanup of stale uploads. R2 orphans removed after 30 days, with a one-line audit log.

None of these were special requests. I told the agent the rule once, in plain English. It built the cron, wired the alert, and pinged me when it was done.

The 4 things that page me

I only get a phone notification for four things:

1. 5xx error rate > 1% for 5 minutes. Almost never fires.
2. Stripe webhook failed 3x in a row. Critical, billing breaks fast.
3. Database storage > 80% of plan limit. Buys me a week.
4. SSL cert expiring in <7 days. Cloudflare-managed, basically never fires, but it's a free check.

That's it. No email storm. No dashboard I have to remember to check. The boring stuff is suppressed unless it's actually wrong.

If your ops setup pages you for everything, you stop reading the pages. The job is to page you only for the things that need a human.

The "small feature" loop

Clients ask for things. They used to text me. Now they email a single shared address — and the agent triages.

The triage is simple:

• Bug → agent investigates, drafts a fix, asks me to review
• Tiny feature (under 4 hours) → agent drafts the change, I review, we ship
• Real feature → agent files it as a TODO, I quote it in next month's call
• Out of scope → agent drafts a polite "this is custom work, can we schedule a scope call"

I review every single agent action before deploy. The agent doesn't push to prod without me. That's a hard rule.

The numbers that matter

The cap is honest: I'm not pretending I work zero hours. I work 6 hours a week on these three. The other 30 hours of my week go to two new projects on fixed-price scopes, where the rate is even better.

What I'd do differently

1. Set the morning report up day 1. I added it for Client A in month 4 and immediately wondered why I'd waited.
2. Get every client on the same alerting setup. Three different alert channels was a small chaos.
3. Do the weekly client digest from the start. Clients pay retainers happily when they see what they're paying for.
4. Don't accept "fix the demo" tickets after launch. Demos are scoped work. Ops is the retainer.