Magic Cloudlets · vs the field

The comparison nobody else will print.

Supabase, Hasura, Lovable, Bolt, Base44, n8n, Power Apps, LangChain — measured on the five axes CTOs actually screen on. Including the rows where the competition genuinely wins. Honest edges make honest claims.

Five axes Nine platforms No asterisks hidden
The five axes
01 Databases — bring the SQL estate you already run, or marry Postgres
02 Self-hosting — Docker, on-prem, air-gapped — or cloud-only
03 Open source — MIT, Apache-with-a-moat, fair-code, or closed
04 Price — platform tax, seat tax, credit tax — or none
05 Time to market — when does the clock actually stop?
The matrix

Nine platforms. Five axes. One honest table.

Cyan is a clean win, amber is a partial or conditional answer, grey is a no. Pricing is described by model rather than dollar figures — vendors change their pricing pages more often than their engines.

Platform Databases Self-hosting Open source Price model Time to market
Magic Cloudlet SQLite · MySQL/MariaDB · PostgreSQL · SQL Server Full — Docker, on-prem, air-gapped MIT — runtime and all Free self-hosted · managed cloudlets from Nuity One sentence → live, validated, RBAC-locked API. Minutes.
Supabase PostgreSQL only Yes — with feature gaps vs the hosted product Apache 2.0 core Free tier + usage-based cloud Instant primitives — the backend logic is still yours to write and review
Hasura Postgres-first; DDN adds SQL Server, MongoDB, ClickHouse Engine yes; DDN tooling gravitates to their cloud Apache 2.0 engines; build tooling cloud-tied Free engine + cloud tiers Instant API layer — business logic lives somewhere else
Lovable Supabase underneath → PostgreSQL No Closed Subscription + credits Demo in minutes — production after the review-and-debug tail
Bolt.new Supabase underneath → PostgreSQL No Closed Credit-based subscription Demo in minutes — production after the review-and-debug tail
Base44 Proprietary built-in database No Closed Subscription tiers Prototype fast — data lock-in on the way out
n8n Not a database — connects to many Yes Fair-code (Sustainable Use License) — not OSI open source Free self-host · execution-priced cloud Fast for glue and automations — not a production API backend
Power Apps Dataverse — proprietary No Closed Per-user / per-app licensing Fast inside Microsoft 365 — a walled garden outside it
LangChain None — it's a library Runs wherever you run it MIT (LangSmith is a paid product) Free library · LangSmith paid A framework — you still build, secure, and host the backend
Assessments reflect publicly documented product behavior as of mid-2026. If we got a cell wrong, tell us — we'll fix it: hello@nuity.ai
Axis 01 · Databases

Bring the databases you already run.

Most of the field gives you exactly one answer: Postgres — either directly (Supabase, Hasura's center of gravity) or inherited invisibly through the vibe coder sitting on top of it (Lovable, Bolt). Base44 and Power Apps go further and put your data in a proprietary store you can't take with you.

A Magic Cloudlet speaks SQLite, MySQL/MariaDB, PostgreSQL, and SQL Server. That last one matters more than the demos admit: mid-size and enterprise companies run SQL Server estates that aren't migrating anywhere. A cloudlet generates a validated backend on top of the database you already own.

Your data model shouldn't be a licensing decision someone else made.
The field Postgres, or a proprietary store

Your existing SQL Server / MySQL estate: migrate it or forget it
Magic Cloudlet SQLite · MySQL/MariaDB · PostgreSQL · SQL Server

Point it at the estate you already run
Axis 02 · Self-hosting

Cloud, your cloud, or no cloud at all.

Lovable, Bolt, Base44, and Power Apps are cloud-only — your backend lives on their infrastructure or it doesn't exist. Supabase and Hasura can be self-hosted, with the familiar catch: the hosted product gets features and tooling the community edition waits for.

Magic Cloud self-hosts fully: Docker, your cloud, on-prem, or air-gapped with no phone-home. The same runtime, the same features, wherever it runs. For regulated industries and data-residency requirements this isn't a preference — it's the qualifying round.

If your compliance team can't put it behind their firewall, it was never a candidate.
Vibe coders & low-code Their cloud or nothing

Data residency: their terms · Air-gap: impossible
Magic Cloudlet Docker · your cloud · on-prem · air-gapped

Full feature set, no hosted-vs-CE gap
Axis 03 · Open source

MIT. Not Apache-with-a-moat. Not fair-code. Not closed.

The licensing spectrum here is wider than the marketing suggests. The vibe coders and Power Apps are simply closed. n8n is "fair-code" under a Sustainable Use License — source-available, but not open source by the OSI definition, with commercial restrictions attached. Supabase and Hasura ship genuinely open engines, with the important parts of the operational tooling pulling you toward their clouds.

Magic Cloud is MIT — the runtime, the whole thing. Audit it, fork it, embed it, sell with it. The license is the anti-lock-in guarantee: if Nuity vanished tomorrow, your backend wouldn't.

A license you have to read twice is a business model, not a guarantee.
The spectrum out there Closed (Lovable, Bolt, Base44, Power Apps)
Fair-code, not OSI (n8n)
Open engine, cloud-tied tooling (Supabase, Hasura)
Magic Cloud MIT — full runtime

No dual license, no usage restrictions, no moat
Axis 04 · Price

No seat tax. No credit meter. No platform rent.

The field monetizes three ways: per-seat (Power Apps — every user is a line item), per-credit (Lovable, Bolt — every generation burns budget), and usage-metered cloud (Supabase, Hasura, n8n cloud). All three grow with your success and none of them cap.

Self-hosted Magic Cloud is MIT-free — the cost is your own infrastructure. Managed cloudlets from Nuity are priced as infrastructure, not per-seat rent. And because generation on a cloudlet collapses auth and hosting into the runtime, the AI token bill for building the backend drops too — we published the math.

Run the numbers on your own team in the savings calculator →
The field Per-seat, per-credit, or usage-metered

Cost scales with adoption — success is billable
Magic Cloudlet MIT self-host: free · managed: infrastructure pricing

Plus ~80% fewer generation tokens per backend
Axis 05 · Time to market

The clock stops at production. Not at the demo.

Here's the twist in this axis: the vibe coders look fastest. They are — to a demo. Then the tail begins: reviewing generated free-form code, chasing hallucinated calls, fixing the auth layer the model improvised. The BaaS platforms and frameworks are honest about it — they hand you fast primitives and leave the actual backend logic to you.

A cloudlet's clock stops later in the sentence and earlier in the calendar: one prompt becomes a live API where every instruction is statically proven to exist and RBAC is enforced by the runtime — so "it works" and "it's safe to ship" happen in the same minute.

Fast to demo is a party trick. Fast to production is a business.
Vibe coders Minutes to demo
Days of review, debugging, and auth hardening to production
Magic Cloudlet Minutes to a validated, role-locked, hosted API
Nothing unproven ever ran
How that proof works → the proof loop, in full
The fine print, up front

Where they're better. Where none of them can follow.

A comparison you can trust has to concede real points. Here are ours.

Genuinely better than us, today

Supabase has a larger ecosystem and community than Magic Cloud. Lovable and Bolt produce prettier frontends faster. Power Apps wins inside an all-Microsoft shop with existing E5 licensing. LangChain offers more orchestration flexibility if you're building the whole stack by hand anyway. If those are your deciding factors, they may be the right call — we'd rather tell you that in a table than have you find out in a migration.

What none of them do

Not one platform on this page statically proves generated logic against the machine's real capabilities before execution, or enforces RBAC at the runtime itself rather than in generated code. That's the axis this table can't show — because every other cell in that column would be empty. See why that's structural, not incidental →

Talk to an engineer, not a sales deck

Bring your current stack. We'll run the comparison live.

20 minutes. Your database, your compliance constraints, your team size — we'll walk this table against your reality and tell you plainly if a cloudlet beats what you have.