8 Supabase Alternatives in 2026 (Open Source and Managed)

The free project paused over a weekend because nobody logged in. Nobody warns you about that one. The invoice grows a compute line, then an egress line, then storage overages. By the second project you are doing arithmetic before every deploy. Some teams need SAML SSO and find that dashboard SSO waits for the $599 monthly Team plan. Any of these sends teams shopping. These are the eight options worth evaluation time in 2026, with databases, licenses, and prices compared.

TL;DR:

• Postgres with branching and scale-to-zero: Neon (free tier, then pay per usage)

• Closest open-source backend platform, BSD license: Appwrite (free cloud, Docker self-host)

• Instant GraphQL over Postgres: Nhost (from $25/mo)

• One binary on a five-dollar VPS: PocketBase (MIT, free forever)

• Mobile apps on the Google stack: Firebase (pay as you go)

• TypeScript-first reactive backend: Convex (free tier, Pro $25 per developer/mo)

• Already committed to AWS: Amplify (per-service pricing)

• A whole app platform with auth, workflows, mail, and payments: Blocks OS (free tier, open-source services)

Why teams search for Supabase alternatives

Three frictions come up again and again. None of them makes Supabase a bad product; it has 114,000 GitHub stars and an Apache-2.0 license for a reason. But they explain the steady traffic to this comparison.

The bill has more moving parts than the brochure suggests. The Pro plan costs $25 per month and includes $10 of compute credits, which covers exactly one project on the smallest Micro instance. Every additional project runs its own compute instance, billed hourly, from $10 to $3,730 a month at the top end. Features that other platforms fold into the plan price sit outside it here. Point-in-time recovery costs $100 per month, a custom domain $10 per project, log drains $60 each. Spend caps help, and they are on by default. The arithmetic is still yours to do.

The free tier pauses projects. After one week of inactivity, a free project goes to sleep, and you are limited to two active projects. For a hobby build that is an annoyance. For an internal tool with sporadic use, or a proof of concept you show a client on Friday and again three weeks later, it is an outage you have to explain.

It is a database platform first. Auth, storage, edge functions, and realtime arrive with every project, and the auto-generated REST and GraphQL APIs over Postgres are good. What surrounds the database in a real product, meaning background workflows, transactional mail, payments, release management, still comes from somewhere else. If you would rather assemble fewer vendors, the last entry in this list exists for exactly that reason.

1. Neon: Postgres with branching and scale-to-zero

Neon, now part of Databricks, is what you pick when the database is the product decision and you want managed Postgres done well. Compute suspends after five minutes of idle time on the free plan and resumes on the next connection, so a quiet dev environment costs nothing. Branching works like git: copy-on-write database branches for every preview deployment or experiment, ten of them free per project.

The platform has grown past plain hosting. Managed Better Auth covers up to 60,000 monthly active users per project on the free tier. Functions, Object Storage at $0.023 per GB-month, and an AI Gateway round out the backend, and the Postgres extensions list includes pgvector, PostGIS, and TimescaleDB on every plan.

Strengths: scale-to-zero that saves real money on non-production databases; branching that fits preview-based workflows; pay-as-you-go with no monthly minimum on the Launch plan.

Weaknesses: the free plan suspends compute after five minutes, which means cold starts on the first query; the broader backend surface (auth, functions, storage) is newer than the database core; no self-hosted option.

Pricing: Free (100 CU-hours, 1 GB storage per project, 10 branches). Launch and Scale are usage-based, from $0.106 per compute-hour.

Best for: teams that want Supabase's Postgres without the platform around it, and startups that branch databases per pull request.

2. Appwrite: the closest open-source equivalent

Appwrite is the alternative that reads most like Supabase: auth, databases, storage, functions, realtime, and messaging in one platform, open source under the BSD-3-Clause license with 57,800 GitHub stars. Self-hosting is a single Docker command. The cloud has a permanent free plan, and the 2026 releases pushed it toward agents: a hosted MCP server lets Cursor or Claude Code operate on a live project, no API keys to manage.

The default database is Appwrite's own TablesDB. Managed PostgreSQL and MySQL are available as dedicated options, with tiers from $10 per month per database, which matters if SQL portability is why you are leaving Supabase in the first place.

Strengths: permissive license with commercial use settled; one-command Docker self-hosting; SOC 2 Type II certification; built-in messaging across email, SMS, and push.

Weaknesses: the free cloud plan freezes projects that hit resource limits until the cycle resets; the document-and-table data model differs from Postgres unless you pay for dedicated SQL; a smaller community than Supabase's.

Pricing: Free plan on cloud, and self-hosting costs nothing. Dedicated database tiers from $10/mo; Pro plans for scale.

Best for: teams that want the self-host-everything option under a license legal already has a folder for.

3. Nhost: instant GraphQL over Postgres

Nhost packages Postgres, the Hasura GraphQL engine, authentication, storage, and serverless functions into one stack. Connect a database and Hasura generates the GraphQL API over it, with realtime subscriptions, event triggers, and role-based authorization at the engine level. Auth includes email and password, OAuth providers, magic links, 2FA, and WebAuthn/FIDO2, which is a shorter list of add-ons than most.

The free Starter plan allows one active project and pauses it after a week of inactivity, so the same caveat as Supabase's free tier applies. Pro starts at $25 per month with $15 of compute credits included, and the Team plan at $599 adds SOC 2 Type II.

Strengths: the fastest route from schema to a production GraphQL API; role-based auth enforced at the engine level, outside your resolvers; infrastructure-as-code through the Nhost CLI.

Weaknesses: the ecosystem is small, 9,400 GitHub stars against Supabase's 114,000, so fewer answers exist to copy when something breaks; the API is GraphQL and nothing else, which will not suit REST shops; compute costs the same arithmetic as Supabase.

Pricing: Starter free (1 project, 1 GB database). Pro from $25/mo plus compute at $15 per shared vCPU core per month or $50 dedicated.

Best for: GraphQL-first teams who consider Hasura a feature and want it managed.

4. PocketBase: an entire backend as one file

PocketBase is a single Go binary with an embedded SQLite database, a realtime subscription layer, authentication with OAuth2, file storage, and an admin dashboard. You drop it on a five-dollar VPS, run it, and you have a backend. The MIT license permits anything, including shipping it inside a commercial product. Extend it with hooks in Go or JavaScript when the built-ins run out.

The trade is scale. PocketBase scales vertically, on a single server, by its own FAQ's description, and it serves that scope well: the project reports 10,000-plus persistent realtime connections on a $4 Hetzner VPS with 2 vCPUs and 4 GB of RAM. Beyond a single machine there is no clustering story, and SQLite concedes heavy concurrent writes. The version number is honest too. v0.40.x sits ahead of a 1.0 that will pin the API down, so upgrades between minor versions can carry breaking changes.

Strengths: the smallest operational surface of anything in this list; MIT license with zero restrictions; the admin UI means non-engineers can manage collections; backups are file copies.

Weaknesses: vertical scaling only, with no clustering story; SQLite concedes heavy concurrent writes; pre-1.0 breaking changes are documented policy; no managed cloud, so uptime is your job.

Pricing: Free, forever, wherever you run it.

Best for: indie products, internal tools, and anything where one server is enough and owning the whole stack matters.

5. Firebase: the incumbent with the deepest free tier

Leaving Supabase for Firebase usually means you either came from there or you build mobile. Firestore and the Realtime Database are document stores with no SQL layer. The free quotas are generous by any standard: 50,000 monthly active users for authentication, 1 GiB of Firestore storage, 2 million Cloud Function invocations per month. Crashlytics, Cloud Messaging, Remote Config, and Analytics cost nothing and have no real Supabase equivalent.

The costs hide in the model. Firestore bills per document read and write, so a chatty screen can cost more than a whole Postgres instance. SAML and OIDC sign-in for enterprise identity lands on the Identity Platform path with 50 free MAUs before per-user billing. The moment you want any of the serious quotas, the project needs a Cloud Billing account, which changes the conversation with finance.

Strengths: the strongest mobile SDKs in the category; free tiers that carry real products; an operations suite (crash reporting, analytics, push, remote config) Supabase does not ship.

Weaknesses: no SQL anywhere, and migrating back to relational later hurts; per-read pricing punishes naive query patterns; costs spread across multiple Google Cloud services that each have their own console.

Pricing: Spark plan free. Blaze is pay as you go, priced per product at Google Cloud rates.

Best for: mobile-first teams, and shops where the rest of the stack is already Google.

6. Convex: the TypeScript-native reactive backend

Convex replaces both the database and the API layer with TypeScript functions backed by a reactive document store. Queries re-run and the UI updates when their underlying data changes, with no websocket plumbing of your own. Crons, webhooks, file storage, vector search, and authentication with 80-plus OAuth integrations are part of the platform, and the code is open source with local development and self-hosting available.

The free tier includes a million function calls before usage pricing starts. Professional costs $25 per developer per month and raises the included usage. Business starts at a $2,500 monthly minimum, so the jump from hobbyist to company scale is steep.

Strengths: end-to-end type safety from schema to UI; reactive queries without hand-built realtime; a development loop that TypeScript teams tend to stick with once they have tried it.

Weaknesses: there is no SQL database, so Postgres skills and tools do not transfer; per-developer pricing differs from every usage-based rival; the query model is its own paradigm, which is a real training cost for a team. Pricing: Free tier with usage beyond included amounts. Professional $25 per developer/mo. Business and Enterprise from $2,500/mo.

Best for: product engineers who live in TypeScript and want the backend to feel like the rest of the codebase.

7. AWS Amplify: when the company already runs on AWS

Amplify is the wrapper that makes assembling a backend from Cognito, AppSync, DynamoDB, Lambda, and S3 feel like one product. Hosting gives you git-push deploys with preview URLs, and the Gen 2 programming model is CDK-based, so what you write is real infrastructure code. Team seats are free, which reads well next to per-seat pricing elsewhere.

The honest description of Amplify pricing is "several invoices". Build minutes, bandwidth, storage, and requests each bill separately, with the free tier easing the first year for new accounts. Estimating a production workload means reading five pricing pages. Cognito's developer experience remains the tax everyone complains about.

Strengths: nothing in the list scales with more headroom, because underneath it is AWS; CDK-based backend definitions that engineers can review; free collaboration seats.

Weaknesses: cost forecasting requires spreadsheet work across services; Cognito auth is workable but widely disliked; you inherit AWS complexity even for small projects.

Pricing: Hosting free within the first-year allowances, build minutes at $0.01 each on the standard instance. Backend services bill at their own AWS rates.

Best for: teams whose security review, billing, and tooling already assume AWS.

8. Blocks OS: the database plus everything around it

Blocks OS is an enterprise backend-as-a-service platform from SELISE, and it takes the opposite bet from Supabase: it starts with the full set of services a business application needs and puts the database platform inside that. There is a free tier ($0/month, up to 10 projects, self-serve signup), and the services themselves are open source on GitHub, so running them on your own infrastructure is allowed even though no packaged self-hosting guide exists yet.

The Data Gateway works the other way around from raw Postgres. You define typed schemas, and the gateway generates per-item CRUD and GraphQL over them, backed by a managed database or by an external database you map in. Field-level validation and access policies live next to the schema, so a write the rules reject stops at the gateway.

Around data sit the services you would otherwise assemble from vendors. IAM handles hosted login with MFA, organizations, and role-based access, including RFC 8693 token exchange so background workers act with a user's context. Logic is a visual workflow canvas with webhook and email triggers, HTTP and database actions, and an AI agent node. Utilities covers transactional mail, hosted payments with Adyen and Stripe webhooks and reconciliation, notifications, PDF generation, and magic links. Release does git-based deployment with SAST and SCA scanning on every build, and there is one CLI and one SDK for all of it:

npm install -g @seliseblocks/cli-os@latest
blocks data schema list --json

const products = blocksClient.data.collection("Product");
await products.create({ title: "Lawn chair", price: 8900 });
const page = await products.list({ pageNo: 1, pageSize: 20 });

Strengths: one platform for data, identity, workflows, mail, payments, and deployment; a free tier that includes auth, AI agents, localization, email, logs, and deployment automation; access control enforced at the gateway, before application code runs; the services are open source, which takes some of the lock-in fear off the table.

Weaknesses: no packaged self-hosting story yet, so "we run it ourselves" means working from the repositories; a younger community and thinner tutorial ecosystem than a 114,000-star project; the enterprise features and support come through a sales conversation.

Best for: teams building business applications that would otherwise glue together five vendors and five invoices. The Vercel vs Blocks OS comparison covers the platform bet in more detail.

Comparison table

Which one should you pick

Start from the constraint, because the feature lists all blur together after a while.

If money is the constraint, PocketBase on a small VPS is unbeatable at the low end, and Neon's free tier carries serious Postgres work at zero. If the license is the constraint, Appwrite's BSD-3-Clause and PocketBase's MIT both permit embedding and resale without a conversation with legal; Supabase's own Apache-2.0 does too, which is worth remembering before you migrate anything. If GraphQL is the constraint, Nhost. If TypeScript is the religion, Convex. Mobile teams should shortlist Firebase before anything else, and AWS shops should look at Amplify before fighting their own procurement team.

And if the actual project is a business application, with users and roles and workflows and invoices, count the vendors in your planned architecture first. Five integrations is five invoices, five failure modes, and five support threads. That is the specific problem Blocks OS was built to solve, and the Auth0 alternatives list plus the n8n alternatives comparison show the same pattern in the identity and automation categories.

Staying on Supabase is a legitimate answer, to be clear. It self-hosts under Apache-2.0, and the Postgres core is standard enough that psql and pg_dump work against any project. The 114,000 stars bought you something: answers. Every question you will ever have already has a Discord thread. The alternatives above are for the teams whose constraint Supabase does not fit.

FAQ

Is Supabase still free to self-host? Yes. The Docker setup and the Supabase CLI are free under the Apache-2.0 license. Only the managed cloud has paid plans, and its free tier pauses projects after one week of inactivity.

What is the cheapest Supabase alternative? PocketBase, by a distance. The MIT-licensed binary runs on a five-dollar VPS with no per-request billing. On managed infrastructure, Neon's free tier and Appwrite's free cloud plan are the strongest zero-cost options.

Which alternative keeps me on Postgres? Neon and Nhost. Neon treats the database as the product, with branching and scale-to-zero; Nhost wraps Postgres in Hasura's instant GraphQL. Appwrite offers managed Postgres as a dedicated option from $10 per month.

Can I move my database off Supabase? Yes, because every Supabase project is a standard Postgres instance with a normal connection string. Any Postgres client, ORM, pg_dump, or psql works against it, so the dump-and-restore path into any other Postgres host is ordinary database work.

Pricing and plan limits verified from each vendor's own pricing page on October 26, 2026. Star counts drift; treat them as approximate.