Deploy & Maintenance

Shipping it, and keeping it healthy after. A project isn't done when it deploys — it's done when I can watch it run, hear about problems, and fix them without drama.

Overview

This phase covers the last mile and everything after: validating the build, releasing it, deploying to a platform with automatic previews, then wiring observability so production tells me when something's wrong. Maintenance is the long tail — and it's easier when monitoring was set up on day one.

Goals

• Validate the build before it reaches users.
• Release and deploy with confidence and easy rollback.
• See real behavior and catch errors in production.
• Keep the app healthy over time with low operational overhead.

Activities

Tools & skills

• Vercel for hosting, previews and web analytics.
• Sentry for error and performance monitoring.
• Microsoft Clarity for heatmaps and session replay.
• The Vercel MCP to deploy and read logs from the agent.
• The PR Review & Commit prompt to wrap up each change cleanly.

Deliverables

• A validated, released build live on a real URL.
• Error and performance monitoring in place.
• Product analytics collecting real usage.
• A maintenance loop: triage, fix, update.

How Morpha UI did it

The only "CI" this project has is npm run types:check (MDX/Next type generation plus tsc --noEmit), and I run it as the gate before anything ships — it's the validation step here. The site deploys to Vercel, where preview URLs let me verify each step before promoting. The Vercel MCP means the agent can deploy and read runtime logs directly, closing the loop from "ship" to "diagnose" without leaving the terminal.

Checklist

• The build is validated (type-check / CI green) before release.
• Deployed via preview URL first, then promoted to production.
• Error and performance monitoring is live.
• Product analytics is collecting usage.
• There's a real loop for triaging and shipping fixes.