When you turn on public access for a database, its hostname now starts with the database's name, for example orders-api-7k2m9q.us-east-1.db.computesphere.com.
Hostnames are public: they appear in DNS, so don't put anything sensitive in a database's name.
Existing hostnames don't change. See PostgreSQL: public endpoint.
]]>SphereCache now goes up to Cache 10 GB.
Available in Sweden and US East. Prices for every size are on the pricing page.
]]>You can now add a managed Valkey cache to any environment, next to your services, at a fixed monthly price.
It runs Valkey, the open-source in-memory data store from the Linux Foundation. Valkey client libraries and most existing cache clients connect without changes.
rediss://), on a private address that only services in the same environment can reach.Available in Sweden and US East. Prices for every size are on the pricing page.
]]>You can now add a RabbitMQ message broker of your own to any environment, next to your services, at a fixed monthly price.
A RabbitMQ 4.3 broker of your own, for background jobs, work queues and events between services. Any AMQP client works, Celery included.
amqps://, port 5671), on a private address that only services in the same environment can reach.Available in Sweden and US East. Prices for every size are on the pricing page.
]]>Create managed PostgreSQL databases in EU (Sweden), eu-north-1. It's open to every account, on every plan from Free to Plus.
csph db instance-types lists each one with its monthly price for your environment's region.DATABASE_URL is injected for you.csph db connect and csph db tunnel reach the database from your machine. Turn on the public endpoint for BI tools: only addresses on its allowlist can connect, and TLS is verified.csph environment create --name prod --project <project-id> --region eu-north-1
csph db instance-types --environment <environment-id>
csph db create orders --environment <environment-id> --instance-type "Standard 1"
Use csph 0.18.2 or later.
]]>Choose which variables your build can read, run migrations before a release takes traffic, and ship the exact image you tested in one environment to another without rebuilding it.
Every variable and secret now has a scope: Build & runtime, Runtime only, or Build only. Plain variables default to Build & runtime; secrets default to Runtime only and reach the build only when you opt them in.
| Build type | How build-scoped values arrive |
|---|---|
| From source, no Dockerfile | Environment variables of your build command, any name. Handed over as a build secret, so they aren't written into the image. |
| From your Dockerfile | Plain variables as build args (declare them with ARG); secrets as build secrets at /run/secrets/<KEY> via RUN --mount=type=secret. |
The console flags browser-bundle variables (NEXT_PUBLIC_*, VITE_*, and friends) as visible to anyone, warns when a secret is added to the build, and offers Save & rebuild when a change only a new build can apply. Read the guide →
Set pre_deploy_command on a service's build manifest and it runs once per build, in the new image, with the service's variables and secrets and access to its datastores — after the build, before any traffic moves. Exit 0 and the rollout proceeds; anything else fails the build and the current version keeps serving. 30-minute limit, output on the build in the Builds tab. Read the guide →
Deploy the image running in one environment to another — dev to production, or any pair you use — by digest, with no rebuild. The target keeps its own variables, secrets, and pre-deploy command. If a value baked into the image at build time differs from the target's, the promotion stops and names the variables so you can decide. Available from Promote to… in the console and POST /v2/deployments/{id}/promote. Read the guide →
One sidebar, fourteen sections, organized around what you're trying to do — deploy a service, manage data, expose a domain, watch logs. Replaces the previous tab-strip of six sections.
The docs site used to split navigation across six top-level tabs (Getting Started, Features, Console, Platform, Support, References). You'd jump between sidebars depending on which tab you landed on, and a single feature could appear in two places with subtly different content.
Now there's one sidebar that travels with you everywhere on computesphere.com/docs. Top-level sections track the customer journey:
URLs didn't change. Bookmarks and AI-agent links keep working.
Premium dev-tool docs (Render, Vercel, Linear) all use a single task-first sidebar. We've followed suit. Skimming for "how do I add a custom domain?" should not require knowing which tab the author filed it under.
]]>Every error returned by the ComputeSphere API now ships with a stable code and a direct link to its catalog entry — paste the URL into Claude, Cursor, or ChatGPT and they have full context for debugging.
Every error response carries the same shape:
{
"type": "https://docs.computesphere.com/errors/trial_expired",
"code": "trial_expired",
"title": "Trial period ended",
"status": 402,
"detail": "Your 14-day trial has ended and no active subscription is on file.",
"request_id": "f47ac10b-58cc-4372-a567-0e02b2c3d479"
}
Switch on code in your client — title and detail may evolve as we improve copy, code will not.
The full catalog lives at /errors. Each entry includes plain-English explanation, action-led fix steps, retry guidance, and code samples in TypeScript, Python, Go, and curl.
You can jump straight to any error page by appending the code: https://docs.computesphere.com/errors/<CODE>.
When something fails at 2 AM, the difference between "an opaque HTTP 500" and "a page that names the cause and shows the fix" is the difference between staying online and waiting on support.
]]>New signups must verify their email address before creating their first project. Existing accounts are unaffected.
After signing up, you'll see a banner in the Console with a "Resend verification email" button. Clicking the verification link in your inbox unlocks the rest of the account immediately — across every open tab. You can still explore the Console while verification is pending; project and service creation just waits for the link.
If you call the API before verifying, you'll get a clear email_verification_required error (HTTP 409) with a link to the docs.
Cuts down on signup spam and accidental typos in email addresses — and means support tickets reach the right inbox the first time.
]]>Every paid plan now starts with a 14-day free trial. Add a card at signup, run your services on us for two weeks, first charge fires on day 15. Cancel anytime in the Console — no charge if you cancel before the trial ends.
Trials should feel like trials, not gauntlets. Two weeks is enough to ship a real thing, send it to a teammate, and decide whether to keep going.
]]>The new v2 API is live at api.computesphere.com/v2. Predictable response envelope, stable error codes, and a published OpenAPI spec that's the source of truth for every client we ship.
{object, items|fields, pagination} for lists, {object, …fields} for single resources/api-reference — browse interactively or download the YAML for your own toolingThe legacy v1 API stays online — no deadline, no surprise removal. New endpoints land on v2 first; a migration guide is on the way.
v1 grew with us — each endpoint shipped its own envelope, error shape, and pagination dialect. v2 is the version we should have shipped first: one envelope, one error shape, no surprises.
The Quickstart is rebuilt around three CLI commands. From csph init to a live service URL on computesphere.app in about a minute, with one-click "Add to Claude" and "Add to Cursor" handoffs for the agent-driven flow.
csph init — scaffold a new service in the current directorycsph deploy — push the code, build the image, run the servicecsph open — open the live URL in your browserThat's it. The page also has an Agent Toolbelt at the top — copy the docs into an AI coding tool with one click so it has full context.
The first impression of a platform is the time between "I want to try this" and "this thing is running." Sixty seconds. Three commands. Two clicks for the agent path.
]]>Verify your email in one tab and the other six update instantly. Switch accounts in one tab and the rest follow. No more stale states, no more "log out and back in to refresh."
Most people working with cloud infrastructure have at least three tabs open at any moment — Services, Logs, Billing. Drifting states between them was a constant low-grade annoyance. Now they all agree on what's true.
]]>Every empty page in the Console — no services yet, no alerts, no logs — now opens with a clear next step instead of a blank screen.
We swept the Console for every "you don't have any X yet" state and gave each one:
Empty states are the most underrated page on any product. They're the first thing a new account sees, and the only thing an account with zero data ever sees. Making them earn their slot beats every other onboarding tweak combined.
]]>