Appwrite Cloud

What Appwrite Cloud actually is

Appwrite is an open-source backend-as-a-service: database, authentication, file storage, serverless functions, messaging, realtime subscriptions, and hosting, exposed through REST APIs and SDKs for web, mobile, and server languages. The project launched publicly in 2019, built by founder Eldad Fux out of the Israeli open-source community, and the core repository runs under a BSD 3-Clause license with roughly 57,000 GitHub stars. Appwrite Cloud is the managed version of that same software, in public beta since 2023 — you can let Appwrite run it, or run it yourself on any host that can execute a Docker command.

That dual identity is the whole pitch, and it is what separates Appwrite from the Firebase-shaped competition. Supabase and Firebase sell you a managed backend and a data-gravity situation; Appwrite sells the same convenience with a documented exit. The database layer, the auth system, the storage API — all of it is code you can lift onto your own infrastructure if the pricing, the roadmap, or the company stops suiting you. In our view that escape hatch is worth more than any single feature comparison, and it is why we take the project seriously despite an ecosystem smaller than the category leaders'.

The feature set is genuinely complete for a small-to-medium application: multiple database types (a TablesDB default, with DocumentsDB, VectorsDB, PostgreSQL, and MySQL available on dedicated compute tiers), user auth with the usual providers and MFA, file storage with permissions, functions with build pipelines, messaging channels, and realtime with subscriptions. MCP support is now included so coding agents can operate on a live project — a detail that dates the current positioning but reflects where the tooling is heading.

Pricing: what you really pay

Appwrite changed its pricing materially in September 2025, and the current structure is per-project rather than per-seat — a change that made some things cheaper and others more expensive depending on your shape. The Free plan remains $0: 5 GB of bandwidth, 2 GB of storage, 750,000 function executions, 75,000 monthly active users, one database, one bucket, and two functions per project, with 15-minute build timeouts. Two hard constraints come with it: a limit of two projects, and free projects are paused after one week of inactivity. For a side project you open every weekend that pause is a mild annoyance; for anything real it is the signal to upgrade.

Pro starts at $25 per month per project, with dedicated resources per project: 2 TB of bandwidth (up from 300 GB before the pricing change — a 6.7x increase, and the single best thing in the new structure), 150 GB of storage, 3.5 million executions, 200,000 monthly active users, unlimited databases, buckets and functions, daily backups kept for seven days, email support, and $10 of monthly compute credits for database usage. Overage is pay-as-you-go: additional bandwidth at $15 per 100 GB (down 62% from the old $40), additional storage at $2.8 per 100 GB, function GB-hours at $0.06. Realtime is metered too — 6 million messages per month on Pro, then $2.50 per million, with 500 concurrent connections included and extra connections at $5 per 1,000. Budget caps exist and are documented, which we always like to see.

Our reading: the bandwidth economics are now genuinely competitive — 2 TB on a $25 plan is generous next to most PaaS — but the per-project resource pool means an architecture with several projects multiplies the base fee ($15 per additional project after the first). Teams that split staging and production into separate projects should model that from day one. Dedicated database compute starts at $10 per month per database, and HA replicas add 50% of the base per replica, point-in-time recovery adds 20%.

Performance & infrastructure

Appwrite Cloud runs in six locations across five countries: Frankfurt, Toronto, New York, San Francisco, Sydney, and Singapore. That is a reasonable spread — North America, Europe, and APAC each get a region — though single-city coverage per continent means European users outside the Frankfurt catchment and APAC users far from Singapore will see latency. For a product whose database is the request path for most API calls, region choice matters more here than for a static host; put your project in the region closest to your users, not your office.

The architecture is serverless by default on Free and Pro, with dedicated reserved compute as an option — the same project can move from serverless quota-based operation to dedicated CPU and memory tiers without a rewrite, and TablesDB's default keeps you out of document-model gymnastics unless you ask for them. Realtime connections count against general bandwidth, which is an important detail for chat-like features: the 250 (Free) / 500 (Pro) concurrent connection caps and the message allowances are the practical ceilings for live features, and the changelog documenting these limits is written with unusual clarity.

Reliability is the harder question to answer from the outside. Appwrite Cloud is still officially in beta-era maturity signals (the pricing page, the changelog cadence, the community-weighted support model), there is no published uptime SLA on the pricing or product pages, and the company is venture-funded and small relative to the hyperscalers. The self-hosted option is the mitigation: your own Postgres, your own backups, your own uptime story. We have not benchmarked Appwrite — there is no honest way to map our server benchmark suite to a BaaS API — so all performance statements here are judgment from documentation and public reporting.

Where it falls short

Ecosystem depth, first. Firebase and Supabase have more integrations, more third-party tutorials, more libraries, and more people who have solved your exact problem in public. Appwrite's community is real and active, but smaller; you will read source code more often than blog posts. That is fine for experienced teams and friction for beginners.

Second, the fine print on the free tier. Pausing projects after a week of inactivity and the two-project limit are reasonable commercial choices, but they make the free tier a trial environment rather than a permanent home — unlike alwaysdata's free plan, say, which is built to last. Budget for Pro.

Third, pricing complexity is growing. The September 2025 change improved bandwidth dramatically but introduced per-project fees, compute credits, separate dedicated tiers, HA replica surcharges, and realtime metering. None of it is hidden — the changelog is candid — but the "predictable pricing" the platform once leaned on now requires a spreadsheet for any non-trivial architecture. The base increase from $15 to $25 per month also hit existing customers directly.

Fourth, operational maturity. No published SLA, community-first support on the lower tiers, and a managed platform that is younger than its GitHub repository suggests. Mission-critical workloads need either the Scale/Enterprise path or self-hosting with your own operations.

Worth noting for teams already mid-migration: the pricing change also moved resources from organization-level to project-level, so each project gets its own dedicated pool instead of sharing an organization quota. That is better isolation and easier cost attribution, at the price of the per-project base fee. It is a coherent trade — just make sure your architecture's project count is a deliberate number before the first invoice.

Who should pick this

Our pick for developers and small teams who want a complete managed backend — auth, database, storage, functions, realtime — but refuse to accept lock-in as the price of convenience. It is a strong fit for agencies shipping client apps, startups validating products quickly, and any team that likes the Firebase model but wants an open-source, self-hostable foundation underneath. If data sovereignty is a future possibility, the ability to move the identical stack onto your own Docker host is the deciding factor.

Do not pick it if you need a published uptime SLA, want the deepest possible integration ecosystem, or are building something where a week of project inactivity would matter on the free tier and you are not ready to pay. For pure static or container hosting, /category/serverless-platforms alternatives with simpler bills will serve you better; for pure managed Postgres, look at the /category/database-hosting picks.

Key takeaways

• Open-source BaaS (BSD 3-Clause) with a genuine self-hosting exit — the anti-lock-in position is real, not marketing.
• Pro at $25/month per project now includes 2 TB bandwidth, 150 GB storage, 3.5M executions, 200K MAU; extra bandwidth $15 per 100 GB.
• Free tier (2 GB storage, 5 GB bandwidth) pauses projects after a week of inactivity and caps at two projects — treat it as a trial.
• Six cloud regions (Frankfurt, Toronto, New York, San Francisco, Sydney, Singapore); realtime capped at 500 concurrent connections on Pro.
• No published uptime SLA and a smaller ecosystem than Firebase or Supabase; strongest for teams who value the exit option.

Sources

• https://appwrite.io/pricing
• https://appwrite.io/blog/post/appwrite-pricing-update
• https://appwrite.io/company
• https://github.com/appwrite/appwrite
• https://cloudspress.com/cloud-providers/appwrite
• https://appwrite.io/changelog/entry/2026-03-19