site: /compute claims pass with the entity line, an FAQ and compare links

apps/site/scripts/seo-page-metadata.test.ts

@@ -59,8 +59,12 @@ const PAGES: PageCopy[] = [
   {
     path: "/compute",
     title: "Prisma Compute | Deploy TypeScript Apps and AI Agents on Bun",
+    // Changed on purpose in the 2026-09-30 claims pass: the audit copy said
+    // "long-lived" and "long-running", which the Compute docs do not back. The
+    // description is now the positioning sentence shared with the homepage,
+    // /postgres and the docs Compute index, character for character.
     description:
-      "Prisma Compute deploys TypeScript apps, APIs, and AI agents from your repo as long-lived Bun processes next to Prisma Postgres, with long-running requests and streaming. One platform for your app and its database.",
+      "Prisma Compute hosts TypeScript apps (Node.js, Bun or Next.js) next to Prisma Postgres on one plan. Generally available since August 2026. Free plan, no credit card. Any Postgres client works; Prisma ORM is optional.",
   },
   {
     path: "/pricing",

apps/site/src/app/compute/page.tsx

@@ -5,19 +5,26 @@ import { ProductFeatures } from "@/components/product/product-features";
 import { ProductHero } from "@/components/product/product-hero";
 import { ProductPlatform } from "@/components/product/product-platform";
 import { ProductProblem } from "@/components/product/product-problem";
+import { CompareLinks } from "@/components/sections/compare-links";
+import { Faq } from "@/components/sections/faq";
 
 export const metadata = createPageMetadata({
   title: "Prisma Compute | Deploy TypeScript Apps and AI Agents on Bun",
+  // The entity wording shared with the homepage, /postgres and the docs, kept
+  // verbatim. lib/markdown-pages.ts repeats it for the .md rendition.
   description:
-    "Prisma Compute deploys TypeScript apps, APIs, and AI agents from your repo as long-lived Bun processes next to Prisma Postgres, with long-running requests and streaming. One platform for your app and its database.",
+    "Prisma Compute hosts TypeScript apps (Node.js, Bun or Next.js) next to Prisma Postgres on one plan. Generally available since August 2026. Free plan, no credit card. Any Postgres client works; Prisma ORM is optional.",
   path: "/compute",
   ogKicker: "Prisma Compute",
   ogAccent: "red",
 });
 
 // /compute composes the product sections directly, like /orm: V4 gives this page
-// no testimonial section, so it can't go through ProductPage. It adds no extra
-// sections either, so this is the template's plain shape minus the testimonials.
+// no testimonial section, so it can't go through ProductPage. Apart from that
+// it is the template's shape, including the two optional sections ProductPage
+// renders for /postgres: the shared "Compare Prisma" list and a visible FAQ,
+// both read from computeContent so lib/markdown/product.ts prints them in the
+// same order.
 export default function ComputePage() {
   return (
     <>
@@ -29,6 +36,8 @@ export default function ComputePage() {
       <ProductProblem problem={computeContent.problem} />
       <ProductFeatures features={computeContent.features} />
       <ProductPlatform platform={computeContent.platform} />
+      {computeContent.compare ? <CompareLinks intro={computeContent.compare.intro} /> : null}
+      {computeContent.faq ? <Faq items={computeContent.faq} /> : null}
       <ProductCta cta={computeContent.cta} />
     </>
   );

apps/site/src/components/product/content/compute.ts

@@ -1,7 +1,9 @@
+import { siteConfig } from "@/lib/config";
 import type { ProductPageContent } from "../types";
 
 // /compute copy, from the approved V4 of the Notion card "Product Page Batch
-// One (3-5) - Copy", with the 2026-08-06 client review's changes marked inline.
+// One (3-5) - Copy", with the 2026-08-06 client review's changes marked inline
+// and the 2026-09-30 claims pass applied on top (see the second note).
 //
 // The review asked this page for three things, and the hero tour answers all
 // three: show the deployment experience immediately (Connect → Deploy → Live),
@@ -11,71 +13,103 @@ import type { ProductPageContent } from "../types";
 // Three things V4 deliberately does on this page:
 //  - No testimonial section. /orm and /postgres carry one; Compute does not,
 //    which is why this page composes its sections directly.
-//  - No "in public beta" callout. V3 had one and V4 removed it — don't
-//    reintroduce it from the homepage, which still labels Compute as beta.
-//  - Five features (V4 shipped four; Object Store buckets, launched 2026-07-24,
-//    were added after the review), rendered three-then-two with the last pair
-//    centred — see product-features.tsx.
+//  - No "in public beta" callout. V3 had one and V4 removed it. Compute has
+//    been generally available since August 2026, and the microline says so.
+//  - Four features, rendered two by two (see product-features.tsx). V4 shipped
+//    four; Object Store buckets, launched 2026-07-24, were added after the
+//    review, and the cron card left in the claims pass.
+//
+// Claims pass, 2026-09-30: every product sentence in this file now has a line
+// in apps/docs behind it. The shared platform-stack.tsx section, which /orm,
+// /postgres and /compute all render, is not part of this pass. What went, and
+// why:
+//  - "the easiest way to host", "microsecond", "same machine", "no code
+//    changes": superlatives and latency figures nothing in the docs states.
+//  - "every PR", "branched from production", "branched together": previews are
+//    per pushed branch and need deploy on push set up first
+//    (compute/deploy-on-push.mdx), and production data is not copied into them
+//    (compute/limitations.mdx, "Databases and migrations").
+//  - "deploy it as it is", "long-running", "durable memory": an existing app
+//    adds a Composer declaration first (composer/porting-an-app.mdx), and the
+//    keep-awake primitives are best-effort, not durable
+//    (compute/keeping-instances-awake.mdx).
+//  - "low-latency queries" and "skip the cross-vendor network hop": the softer
+//    cousins of the latency claim. The docs say only that the app runs next
+//    to Prisma Postgres (compute/index.mdx), so the page says that.
+//  - The cron card: cron scheduling is not part of this release and there is
+//    no committed Compute config file (compute/limitations.mdx, "Runtime" and
+//    "CLI"). It returns only as a Composer scheduled-jobs card once the
+//    product team confirms one; the configJobs illustration stays for that.
 const CONSOLE = "https://console.prisma.io/sign-up";
 const DOCS = "/docs";
 
+// The trust line the brief puts on the homepage and /compute. It is built from
+// every entry of siteConfig.proof, the way hero-home.tsx and
+// lib/markdown/home.ts build the homepage's proof line, so the two pages
+// cannot drift apart and nothing here depends on how many entries the list
+// has.
+const TRUST_LINE = siteConfig.proof
+  .map(({ stat, label }, i) => `${i === 0 ? "Trusted by " : ""}${stat} ${label}`)
+  .join(" · ");
+
 export const computeContent: ProductPageContent = {
   name: "Prisma Compute",
   accent: "compute",
   hero: {
     headline: "One platform for your app and its database",
     headlineEmphasis: "One platform",
-    // V4's second sentence is dropped from the hero — it pushed the CTA down,
-    // and the tour beside this copy now demonstrates both halves of it.
+    // The first sentence of the entity wording shared with the homepage,
+    // /postgres and the docs. The microline carries its second and third
+    // sentences; the page's meta description (app/compute/page.tsx) carries
+    // all four verbatim.
     subheadline:
-      "Prisma Compute is the easiest way to host TypeScript apps, with Postgres and the rest of your infrastructure built in.",
+      "Prisma Compute hosts TypeScript apps (Node.js, Bun or Next.js) next to Prisma Postgres on one plan.",
     benefits: [
-      "Every PR gets a dedicated app and database, branched together",
-      "Microsecond app-to-database latency, no cross-vendor network hop",
-      "Long-running, streaming workloads built for hosting agents, not just apps",
+      "Set up deploy on push, and every pushed branch gets a preview with its own database",
+      "Your app runs next to Prisma Postgres, in one of six regions",
+      "Streams responses and scales to zero when idle, for agents as well as apps",
     ],
     primaryCta: { label: "Get started free", href: CONSOLE },
     secondaryCta: { label: "Read the docs", href: DOCS },
-    // Deliberately not "free while in public beta" — see the note at the top
-    // of this file. This says the same encouraging thing without the label.
-    microline: "Deploy from GitHub or the CLI. Free tier, no credit card.",
+    microline: "Generally available since August 2026. Free plan, no credit card.",
     tour: [
       {
         label: "Connect",
         caption:
-          "Point Compute at a GitHub repo, or deploy straight from your machine with one command.",
+          "Deploy from your machine with npx prisma deploy module.ts, or connect the GitHub repo and add the deploy workflow to deploy on every push.",
         illustration: "repoConnect",
       },
       {
         label: "Deploy",
         caption:
-          "One deploy ships the app and its branched database together, migrations included.",
+          "One deploy provisions the services and databases declared in module.ts, and applies the migrations of any database typed by a Prisma ORM contract.",
         illustration: "deployLog",
       },
       {
         label: "Live",
         caption:
-          "Production and a per-PR preview, each with its own app and its own isolated database.",
+          "Production, plus a preview per pushed branch with its own services and database once deploy on push is set up.",
         illustration: "deployments",
       },
       {
         label: "Apps",
-        caption: "Start from an app — Next.js, Hono, an API, an agent — and deploy it as it is.",
+        caption:
+          "Bring a Next.js, Hono or Bun app, or an agent: add a Composer declaration (service.ts and module.ts) around the server code you already have, then deploy.",
         illustration: "runApps",
       },
     ],
   },
   problem: {
     headline: "One platform can ship features two never could",
     body: [
-      "When your hosting provider and your database provider are separate companies, the things that depend on them being aware of each other (branching, preview environments, low-latency queries, end-to-end test data) don't exist. You wire them together yourself, or you do without.",
+      "When your hosting provider and your database provider are separate companies, the things that depend on them being aware of each other (branching, preview environments, end-to-end test data) don't exist. You wire them together yourself, or you do without.",
       "Prisma Compute exists to change that.",
     ],
     outcomes: [
-      { icon: "gitBranch", label: "Branch your stack per PR" },
-      { icon: "rocket", label: "Skip the cross-vendor network hop" },
+      { icon: "gitBranch", label: "A preview per pushed branch" },
+      { icon: "rocket", label: "Ship app and database in one deploy" },
       { icon: "bot", label: "Host streaming agents natively" },
-      { icon: "settings", label: "One config for app and database" },
+      { icon: "settings", label: "One module.ts for app and database" },
     ],
   },
   features: {
@@ -89,27 +123,21 @@ export const computeContent: ProductPageContent = {
       {
         name: "Built for hosting AI agents",
         description:
-          "Agents need long-lived processes, streaming responses and durable memory. Compute supports all three, on the same platform that runs your data.",
+          "Agents stream their answers and spend most of their time waiting on a model API. Compute streams a response as your service sends it, bills memory and active CPU separately, and scales to zero between tasks.",
         illustration: "agentHosting",
       },
       {
-        name: "Branch your stack per PR",
+        name: "A preview per pushed branch",
         description:
-          "Every preview environment gets a dedicated database, branched from production and tied to that deploy. No shared test DB that passes tests production fails.",
+          "Once deploy on push is set up, every pushed branch gets its own services and its own database, and deleting the branch tears them down. A database typed by a Prisma ORM contract gets your migrations applied on deploy; one declared with rawPostgres() keeps your own migration step. No shared test DB that passes tests production fails.",
         illustration: "branchedStack",
       },
       {
-        name: "Microsecond queries, no code changes",
+        name: "App and database on one platform",
         description:
-          "ORMs make many small queries, and on most stacks each one pays a network round trip between two vendors. Compute runs on the same machine as Prisma Postgres, so that round trip disappears.",
+          "Your app runs next to Prisma Postgres, in the same project and on the same plan, with PgBouncer pooling included. Any Postgres client works, and Prisma ORM is optional.",
         illustration: "coLocated",
       },
-      {
-        name: "Cron and background jobs as a first-class concept",
-        description:
-          "Scheduled execution and background jobs are declared in prisma.config.ts, alongside the rest of your app. No external schedulers, no separate sidecar service.",
-        illustration: "configJobs",
-      },
       {
         name: "S3-compatible file storage built in",
         description:
@@ -120,15 +148,56 @@ export const computeContent: ProductPageContent = {
     ],
   },
   platform: {
-    body: "Prisma Compute is where the stack comes together: your app, your database, and your data model on one platform.",
+    // The second sentence is the exit line shared with /postgres, verbatim.
+    body: "Prisma Compute is where the stack comes together: your app, your database, and your data model on one platform. Standard Postgres underneath: leave with pg_dump; the Composer module is a TypeScript file in your repo, not a runtime.",
+  },
+  // The shared "Compare Prisma" reading list (sections/compare-links.tsx), the
+  // same list /postgres and /pricing show. Only the sentence above it is this
+  // page's own, so it must not repeat theirs.
+  compare: {
+    intro:
+      "How Prisma Compute and Prisma Postgres compare with the other hosting and Postgres options, and where each one fits.",
   },
+  // Visible text only: no FAQ schema, and every answer is a sentence the docs
+  // back (the PR's "facts checked" block names the line for each). Plain
+  // strings, the shape /postgres uses, so lib/markdown/product.ts prints them
+  // as they are; where an answer leans on a docs page it names the page in
+  // words.
+  faq: [
+    {
+      question: "Can I host my existing GitHub repo with a TypeScript frontend and Node API here?",
+      answer:
+        "Yes, when the frontend and the API are TypeScript services built for Node.js, Bun or Next.js. Add a Composer declaration around the server code you already have (a service.ts per service and one module.ts); the porting guide in the docs has a prompt your coding agent can write it from. Run your build, then deploy with npx prisma deploy module.ts. For deploy on push, connect the repository and add the prisma/cloud-deploy-action workflow; from then on every pushed branch gets a preview with its own database.",
+    },
+    {
+      question: "What does Prisma Compute not do?",
+      answer:
+        "Compute does not support WebSocket servers, cron scheduling or a persistent filesystem today, and each service runs in one region, chosen from six. A request has 60 seconds to send its first byte, and a response that has started streaming is not cut off. The runtime is Bun. The Compute FAQ in the docs has the full list of trade-offs.",
+    },
+    {
+      question: "Do I need Prisma ORM?",
+      answer:
+        "No. Your app reaches Prisma Postgres, or any other Postgres, through a connection string, so any Postgres client works. Prisma ORM is optional.",
+    },
+    {
+      question: "Is Prisma Compute related to Prisma Cloud?",
+      answer:
+        "No. Prisma Compute is unrelated to Palo Alto Networks' Prisma Cloud. Prisma makes Prisma ORM, Prisma Postgres and Prisma Compute, and the dashboard is Prisma Console.",
+    },
+    {
+      question: "What does it cost?",
+      answer:
+        "The Free plan includes 1M requests, 360 GB-hours of memory, 4 active vCPU-hours and 10 GB of outbound bandwidth a month for Compute, and 200k operations, 500 MB and 50 databases for Prisma Postgres, with no credit card and no usage billing. Starter is $10 a month base plus metered compute: 5M requests and 1M operations are included, memory ($0.006 per GB-hour), active vCPU ($0.064 per hour) and bandwidth ($0.025 per GB) are metered from the first unit, and you can set a spend limit. An idle app scales to zero. The pricing page lists every plan.",
+    },
+  ],
   cta: {
     headline: "One platform. Both layers. Features only possible because of it.",
-    body: "Compute is part of a TypeScript stack where hosting and the database travel as a unit. Branch them together, deploy them together, debug them together.",
+    body: "Compute is part of a TypeScript stack where hosting and the database travel as a unit: preview them together, deploy them together, debug them together.",
     benefits: [
       "Built to host the agents you're building, not just the apps",
-      "Per-PR branching across app and database",
-      "Co-located queries with no cross-vendor latency",
+      "A preview per pushed branch, app and database together",
+      "App next to its database, on one plan",
+      TRUST_LINE,
     ],
     primaryCta: { label: "Get started free", href: CONSOLE },
     secondaryCta: { label: "Read the docs", href: DOCS },

apps/site/src/components/product/illustrations/agent-hosting.tsx

@@ -1,19 +1,22 @@
 import { Bot, Database, Repeat } from "@/components/icons/forma";
 import { Bar, CardChrome, SurfaceCard } from "./parts";
 
-// "Built for hosting AI agents" — the three things the copy says agents need,
-// running in one place: a long-lived process, a streaming response, and durable
-// memory. The streaming line ends in a caret so it reads as still arriving.
+// "Built for hosting AI agents": the three things the copy says about hosting
+// an agent here. The response streams, memory is held (and billed) while the
+// agent waits on a model API, and the service scales to zero between tasks.
+// The streaming line ends in a caret so it reads as still arriving. Nothing
+// here says "durable": the keep-awake primitives are best-effort
+// (compute/keeping-instances-awake.mdx).
 
 const NEEDS = [
-  { icon: Repeat, label: "long-lived" },
   { icon: Bot, label: "streaming" },
   { icon: Database, label: "memory" },
+  { icon: Repeat, label: "scale to zero" },
 ];
 
 export function AgentHosting() {
   return (
-    <SurfaceCard label="Illustration of an agent hosted on Prisma Compute: a long-lived process streaming a response, with durable memory alongside it">
+    <SurfaceCard label="Illustration of an agent hosted on Prisma Compute: a process streaming a response, holding its memory while it waits on a model API, and scaling to zero between tasks">
       <CardChrome file="agent" />
       <div className="flex flex-1 flex-col justify-center gap-3 px-4 py-3 font-mono text-[0.625rem] leading-none text-foreground">
         <div className="flex flex-wrap items-center gap-1.5">