Dora NodaOn July 17, 2026, a new self-hosted PaaS called OpenShip launched via a 132,000-view X thread, with 4,300 GitHub stars already banked before launch day and a pitch aimed squarely at developers escaping Vercel and Railway bills: git-push deploys, one-click databases, object storage, a built-in mail server with webmail, and an MCP server so AI agents can deploy and roll back — all Apache-2.0, all free forever on a VPS you own.
It is the most aggressive bundling bet a self-hosted platform has made in years. Coolify gives you managed databases; OpenShip adds your mail server to the same box.
So here is the verdict up front, with receipts below: the bundle genuinely saves a small team $20–60/month in service fees at low-to-mid volume — and it does so by concentrating every stateful service you run into one failure domain, on one machine, behind one upgrade path. Whether that trade wins depends entirely on what your 3 a.m. looks like.
| Monthly cost, typical small app | 10k emails/mo | 50k emails/mo | 200k emails/mo |
|---|---|---|---|
| OpenShip self-hosted (VPS + SES relay) | ~$14–27 | ~$18–31 | ~$33–46 |
| Unbundled (VPS + managed Postgres + email API) | ~$36–42 | ~$41–47 | ~$60–110 |
Cash is only half the ledger, though. The other half is operational: backups you have tested, upgrades you have rehearsed, and a mail reputation you now personally own. This post prices both halves.
What is actually in the box
OpenShip is built by Oblien, whose founder previously led open-source at Vercel. The architecture is deliberately boring in the right places: your code builds on your machine, not on the production server, producing an immutable versioned artifact that streams to the target over plain SSH and starts as a fresh container. No agent, no daemon, nothing installed on the box.
The previous version stays warm, so rollback is one click with no rebuild. Domains terminate at an edge layer with free auto-renewing SSL, and the container swap happens with zero downtime.
The stateful side is where the bundle lives. Postgres, Redis, mail, and object storage join your app on an isolated private network — reachable by your containers, never by the internet. Launch materials list one-click stacks for Postgres, Redis, Supabase-style Postgres, and Qdrant alongside MySQL and MongoDB support, plus workers, WebSockets, and storage. You drive it four ways: CLI, web dashboard, desktop app (Mac and Windows), or an MCP server that lets Claude or Cursor agents deploy and roll back infrastructure directly.
Two honest caveats from the launch coverage. First, multi-server clustering and HA are still "coming soon," not shipped — today the unit of OpenShip is one box. Second, the docs are thin and the community is a fraction of Coolify's or Dokploy's, which matters exactly when something breaks at night. Neither is disqualifying for a two-month-old project, but both belong in the price.
The mail bundle, honestly priced
The mail server is OpenShip's headline differentiator, so price it first. You get domains, mailboxes, and a modern webmail client from one panel: no SaaS, no per-seat pricing, every byte stored on your server. But read the fine print admiringly, because it is the correct engineering call: outbound mail "sends through a trusted relay (Amazon SES or any SMTP) so mail lands." OpenShip is not asking you to warm up a fresh VPS IP and fight Gmail's bulk-sender filters from scratch. It stores your mail and lets SES deliver it.
That means the real comparison is not "self-hosted mail vs. $20/month Resend." It is "self-hosted mailboxes + SES relay rates vs. an email API's per-thousand rates," plus the ops you absorb. At current 2026 pricing — SES at $0.10 per 1,000 emails, Resend at $20/month for 50,000, Postmark at $15/month for 10,000, SendGrid around $19.95/month — the math for a typical small app looks like this:
| 10k emails/mo | 50k emails/mo | 200k emails/mo | |
|---|---|---|---|
| OpenShip mail (VPS share ~$5–10 + SES relay) | ~$6–11 | ~$10–15 | ~$25–30 |
| Resend | $0–20 | $20 | ~$80+ |
| Postmark | $15 | ~$55–75 | ~$200+ |
| Raw SES, no mailboxes | $1 | $5 | $20 |
The bundle wins on cash at every volume shown — most decisively against Postmark-style deliverability-first pricing. But notice the last row: if you only need transactional sending (password resets, receipts) and not mailboxes with webmail, calling SES directly is cheaper than either option. OpenShip's mail value is the mailboxes, the webmail, the unlimited domains, and owning the data — not the per-email rate, where it is just reselling you SES economics with extra steps.
And the ops half of the ledger is real. Self-hosted mail lands in spam for a small number of fixable reasons — PTR mismatch, missing or misaligned SPF/DKIM/DMARC, no DMARC reporting — and "fixable" still means you fix them, on your DNS, before Gmail and Yahoo accept your mail. A managed email API does IP warmup, reputation monitoring, and blacklist remediation invisibly.
OpenShip's relay design dodges the worst of deliverability, but mailbox storage growth, webmail upkeep, backup of mail data, and the DNS ceremony for every new domain are yours now. Price an hour of that quarterly and the $20/month savings start looking like minimum wage for a mail admin.
The database bundle, honestly operated
One-click Postgres is table stakes in 2026 — Coolify ships eight first-class databases with scheduled S3 backups, Dokploy automates database backups to external storage — and OpenShip matches that pattern with its one-click stacks on the private network. The honest question was phrased perfectly by this year's self-hosted PaaS roundups: plugin-managed Postgres has backups, but verifying restore on a new node is your job.
Run that checklist against any bundled database, OpenShip's included:
| Operational task | One-click bundle gives you | Still your job |
|---|---|---|
| Provisioning | One click, private network wiring | Sizing the shared box for peak of all services at once |
| Backups | Scheduled snapshots to S3-compatible storage | Retention policy, encryption keys, off-box copies |
| Restore | A restore button | Quarterly restore drills onto a different machine |
| Major-version upgrades (Postgres 16 → 17) | A newer image tag | Downtime plan, extension compat, rollback rehearsal |
| Failure isolation | Containers on one host | The database shares CPU, disk, and fate with mail + apps |
The last row is the one that compounds. On a bundled box you do not size Postgres for Postgres — you size one VPS for the app's traffic peak plus Postgres's checkpoint I/O plus mail storage growth plus the build cache, and you pay for that headroom 24/7.
Unbundled, each service scales (and fails) independently: the mail queue can fill a disk without taking Postgres with it, and a Postgres major upgrade never shares a maintenance window with your deploy pipeline. That independence is what managed Postgres fees actually buy. It is not free anywhere; the bundle just lets you pay in headroom and correlated risk instead of dollars.
None of this is an argument that bundling is wrong. For a side project or an early SaaS with one Postgres and modest mail, the correlated-risk discount is enormous: one box to patch, one dashboard, one login, no integration glue between four vendors' APIs. Just go in knowing the failure drill is the product you declined to buy.
One box, one blast radius
Step back and the pattern is clear: OpenShip collapses four historically separate failure domains — deploy pipeline, database, mail, object storage — onto a single Linux host you operate. Each consolidation removes integration tax (no VPC peering, no cross-service auth, no four-vendor debugging) and adds blast radius (one kernel panic, one full disk, one botched platform upgrade away from everything being down at once).
Upgrade coupling is the sharpest edge. When the mail server and the database are managed by the same control plane as the deploy pipeline, a platform upgrade is a simultaneous upgrade of all three. The rollback story OpenShip advertises — previous app version stays warm, one click back — covers your app, not a Postgres data-directory migration or a mail-store schema change that shipped in the same platform release. Until clustering lands, there is also no second machine to fail over to: "highly available" is a roadmap item, and the roadmap is honest about that.
Compare with the trajectory of the single-box tools OpenShip joins. Coolify grew native multi-server orchestration from one dashboard; Dokploy leans on Docker Swarm across Hetzner VPS instances for clustering. Both learned the same lesson: teams adopt the single box for their first app and need the second machine sooner than they expect. OpenShip's bet is that the bundle is compelling enough to earn that patience — 4,300 pre-launch stars suggest the demand is real — but a team choosing it today should have a written answer to "what happens when we need machine two," because the platform's answer is still "soon."
When to bundle, when to unbundle
This is not a "bundle bad, managed good" sermon. It is a scope question, and the right answer depends on team shape:
| Choose the bundle (OpenShip self-hosted) when… | Choose unbundled (deploy-only platform + managed services) when… |
|---|---|
| One app, one Postgres, modest mail — the headroom math is trivial | Stateful services have independent scaling curves |
| You want one login, one dashboard, one box to patch | Mail deliverability is revenue-critical (you want Postmark's reputation team) |
| Data sovereignty matters (every byte on your hardware) | You need HA/failover this quarter, not "coming soon" |
| Token-efficient agent deploys over MCP appeal to you | Your Postgres holds data you cannot afford to restore-drill never |
There is also a middle path the launch thread underplays: run OpenShip for deploys and app hosting, point outbound mail at SES directly (as it does anyway), and keep Postgres managed until the monthly fee exceeds the cost of your own restore drills. Bundles are not all-or-nothing unless you let them be.
The deeper convergence is worth naming. OpenShip's MCP server — deploys and rollbacks callable by Claude or Cursor agents — is part of a real 2026 shift: infrastructure you can describe to an agent instead of clicking through. Coolify, Dokploy, and the Render-compatible APIs are all drifting toward machine-operable surfaces. Whoever wins the self-hosted PaaS race will win it on agent-operability as much as on price, and OpenShip shipping MCP on day one is the clearest-eyed part of the whole launch.
Bex.co is the open-source, AI-native Render alternative — push a git repo, get a running HTTPS service on machines you own. It deliberately stays out of the bundling race: no managed databases, no mail server, just the deploy layer done well and operable by your agents. Star the repo on GitHub or deploy your first app today.
