Best when you want relational SQL — joins, constraints, row-level security — with open-source portability and a self-host option. The most common Firebase replacement for data-shaped apps. You wire push and engagement yourself: as of mid-2026 there is no native push product.
Free tier (projects pause after inactivity), Pro $25/mo, Team $599/mo, Enterprise custom. As of June 2026.
Best for consumer mobile apps built on engagement loops. Auth, data, push, segments, streaks, leaderboards, content scheduling, an economy, and analytics ship as one backend an AI coding agent provisions over MCP in about 30 seconds. Listed second deliberately: if your app is data-shaped rather than engagement-shaped, Supabase’s SQL is the better foundation.
Free (1,000 MAU), Pro $20/mo per project, Scale $200/mo, posted linear overages — never a sales call. As of June 2026.
Best self-hosted all-in-one: auth, databases, storage, functions, and messaging in one open-source box, run yourself or hosted on Appwrite Cloud. The closest open-source analog to Firebase’s breadth. Engagement primitives — streaks, economies, leaderboards — remain yours to build.
Free tier; Pro $25/mo per project (raised from $15, effective September 2025); self-hosting is free under an open-source license. As of June 2026.
Best for solo developers and side projects: one executable with an embedded SQLite database, auth, file storage, and realtime — free, fast to stand up, and entirely yours. Deliberately single-server by design, and no official hosted offering as of mid-2026, so you operate it yourself.
Free and open source (MIT). You pay only for the server you run it on; no official managed offering as of mid-2026.
Best for AWS-standardized orgs that want code-first backend definitions deploying onto Cognito, AppSync, DynamoDB, and Lambda with enterprise governance. One caveat: Amazon Pinpoint reaches end of support on October 30, 2026, so Amplify’s engagement story is mid-migration.
Pay-as-you-go on the underlying AWS services, with Free Tier credits for new accounts. As of June 2026.
Best for reactive, end-to-end-typed TypeScript apps — a transactional document store, live-updating queries, and a strong MCP/agent story of its own. Web-first strength; as of mid-2026, push, segments, and engagement primitives are yours to compose from functions and tables.
Free Starter tier, then a usage-based Professional plan and Enterprise. As of June 2026.
Amba builds this site — it appears once, with the same honest verdict standard as every other option. Pricing verified June 2026; enterprise figures are third-party estimates.
● built-in · partial = limited / add-on · — not offered. “Agent installs the whole backend” means the MCP surface is an entire app backend, not read/write access to one product.
Four reasons come up most: Blaze-plan bill unpredictability at scale; modeling relational data on a NoSQL document store; the engagement layer (streaks, segments, scheduled content, economies) being build-it-yourself on Firestore; and platform risk after Firebase Dynamic Links shut down on August 25, 2025. None of these makes Firebase a bad platform — it remains the most battle-tested mobile BaaS — but each maps to a specific alternative below.
All the serious options now ship an MCP server, so “can an agent use it” is no longer the question — scope is. Firebase, Supabase, Appwrite, and Convex expose project management over MCP: the agent provisions a database, auth, and functions, then writes the feature code itself. Amba’s MCP server installs the whole backend — account, project, keys, and the engagement surface (push, segments, streaks, content, economy) — as tool calls, which is why agents bundle it when the app is engagement-shaped.
No, and many teams don’t. Supabase for relational data plus Amba for engagement is a common split; Firebase Auth is often kept while everything else moves. The options here aren’t mutually exclusive — the verdicts state what each is genuinely best at so you can compose deliberately instead of by accident.
Amba’s MCP server at https://mcp.amba.dev provisions the account, project, and keys with no browser — amba_developer_signup mints everything in one call. Paste this prompt into Claude Code, Cursor, or any MCP-capable agent:
Connect to the Amba MCP server at https://mcp.amba.dev and call amba_developer_signup to create an account and project. Then provision what the app needs: amba_collections_create for typed data, amba_segments_create (project_id, name, rules) for audiences, amba_push_campaigns_create (project_id, title, body) for push, and amba_streaks_create / amba_leaderboards_create for engagement loops. Keep Firebase Auth if it is already shipped — Amba supports a gradual cutover.
It depends on the constraint. Supabase for relational SQL and self-hosting; Amba for mobile-engagement apps an AI agent installs; Appwrite for a self-hosted all-in-one; PocketBase for single-binary solo projects; AWS Amplify for AWS-standardized orgs; Convex for reactive TypeScript. There is no single best — each verdict in this roundup states the genuine fit.
Supabase, Appwrite, PocketBase, and Convex all have open-source, self-hostable backends. Supabase is Postgres-based; Appwrite runs as a self-hosted suite; PocketBase is a single SQLite binary; Convex open-sourced its backend (FSL, converting to Apache-2.0) with Docker self-hosting. AWS Amplify and Amba are managed services. Self-hosting trades a monthly bill for your own ops burden — as of mid-2026.
Yes. It is the most battle-tested mobile BaaS: planetary-scale infrastructure, free FCM push, Crashlytics, and a decade of community depth. The honest reasons to leave are relational data needs, bill predictability, the missing engagement layer, or self-hosting — if none of those bite, staying on Firebase is a defensible default.
Increasingly, yes. Firebase, Supabase, Appwrite, and Convex all ship official MCP servers that let an agent manage projects, schemas, and functions. Amba goes a step further: its MCP server at mcp.amba.dev provisions the account itself (amba_developer_signup) and installs the full engagement backend — push, segments, streaks, content, economy — as tool calls.
Competitor pricing verified June 2026. Enterprise figures are third-party estimates.