Weekly Vercel deploy health report, written to Notion

I want a weekly written report on how our Vercel deploys are trending, saved as a new page in a Notion database so engineering leadership has a paper trail without anyone hand-compiling it.

Trigger: cron, every Monday at 8am in my local timezone.

Inputs the workflow should accept when I set it up: the Vercel team scope (if any), and the Notion database ID where the report page should be created.

What the agent should do each run:

1. Use Vercel List Projects to get the full project inventory for the configured team or personal account. Paginate through all pages so nothing is missed.

2. Use Vercel List Deployments with a since/until time window covering the last 7 days to pull every deployment across all projects. For each deployment capture: project name, state (READY, ERROR, CANCELED, BUILDING), target (production vs preview), createdAt, buildingAt, ready (so build duration can be computed as ready minus buildingAt in milliseconds), the creator's username, and the git commit metadata (sha, message, branch, author).

3. Make a second Vercel List Deployments call for the prior 7 days as a baseline, capturing the same fields. This is what powers the week-over-week comparison.

From that raw data, compute for the current week: total deploys, production deploys, failure rate per project (ERROR plus CANCELED divided by total), median and p95 build time per project, the top 3 most-frequent deployers across the org, and any projects whose failure rate or p95 build time meaningfully regressed versus the previous 7-day window. Use p95 rather than average since build times have a long tail.

Then use Notion Create a Page with the configured database as the parent. The page should look like this:

• Title property: "Vercel weekly report — week of [Monday's date, e.g. May 26]".

• Body block 1: a short executive summary paragraph in plain English (total deploys this week vs last week, overall failure rate, headline trend in one or two sentences).

• Body block 2: a per-project table with columns for project, total deploys, production deploys, failure rate, median build time, p95 build time, and a delta arrow versus last week for failure rate and p95 build time.

• Body block 3: a "What to watch" section. For each regression, write one or two sentences that explain what changed (e.g. "checkout-web's p95 build time went from 2m10s to 4m05s this week, driven by 6 failed production builds on Wednesday") and why it might matter. Do not just dump numbers.

• Body block 4: a "Top deployers" callout listing the three accounts that shipped the most deploys this week.

Quality rules for the agent: write the watchlist commentary in the voice of an engineering lead, not a stats dashboard. Skip projects with zero deploys this week and last week so the table stays readable. If a project had no deploys in one of the two windows but did in the other, still include it and label the missing week as N/A. If the run fails (Vercel rate limit, Notion permission error), fail loudly with a clear error message and do not create a half-empty page.

Additional information

What does this prompt do?

Pulls every Vercel deploy from the last seven days across all your projects and turns it into a written report.
Calls out which projects got slower, which got flakier, and which had a quiet week, so engineering leadership can read the trend without hand-compiling numbers.
Writes the report straight into a Notion database as a new page, with a title, a short executive summary, a per-project table, and a plain-English watchlist.
Compares this week to last week automatically, so the report explains what changed instead of just dumping totals.

What do I need to use this?

A Vercel account with access to the projects you want included in the report.
A Notion workspace with a database set aside for the weekly reports, shared with your Notion connection so it can write into it.

How can I customize it?

Change when it runs. Monday at 8am is the default, but any weekday or time of day works.
Pick which projects are included. By default it covers every project the account can see, but you can scope it to one team or a short list.
Tune what the watchlist flags. Tighten or loosen the thresholds for what counts as a regression in failure rate or build time.

Frequently asked questions

Where does the report get saved?

As a new page inside a Notion database you choose. Every Monday a fresh page is created with that week's date in the title, so you build a paper trail you can scroll back through.

Do I need a paid Vercel plan?

No. The report works on Hobby, Pro, and Enterprise. A few advanced stats only exist on paid plans, but the core report (deploy counts, failure rate, build time, top deployers) works on every plan.

Can I get this in Slack instead of Notion?

This template writes to Notion because leadership wanted a written paper trail. If you'd rather have a weekly Slack post, swap Notion for Slack when you customize the workflow and the same report can be posted to a channel.

How does it know what counts as a regression?

It pulls the prior seven days as a baseline and compares failure rate and build time project by project. If a project's failure rate jumps or its build time gets noticeably slower, it shows up in the watchlist section with a plain-language explanation.

Will it work if I have a lot of projects?

Yes. It paginates through every project and every deploy on its own. Larger accounts just take a little longer to run.