An open-source Firebase alternative built around a dedicated PostgreSQL database for every project. It composes mature open-source tools—PostgREST for auto-generated REST APIs, GoTrue for JWT auth, Realtime for WebSocket subscriptions, Storage for S3-compatible files, Deno for Edge Functions, and Supavisor for connection pooling—into a cohesive backend platform. It ships with 50+ Postgres extensions including pgvector for AI vector search. 2026 updates include native local development without Docker, Declarative Schemas 2.0, Supabase Compute for long-running services and agents, and an app-specific MCP server that lets end-user agents access data through RLS policies. Pricing: Free ($0, 500MB DB, **paused after 1 week of inactivity**, max 2 active projects), Pro ($25/mo, 1 Micro instance + 8GB disk), Team ($599/mo), Enterprise custom. Compute scales from Micro (~$10/mo) to 16XL (~$3,730/mo). **Key migration note**: `anon` and `service_role` keys are being deprecated by end of 2026—migrate to publishable and secret keys. Backed by $500M Series F (June 2026) at $10.5B valuation, 4M+ registered developers.

Pros & Cons:

Pros: The dedicated Postgres instance per project means you get full database privileges and portability—no lock-in, standard connection strings work with any Postgres client, and pg_dump can export everything; the open-source core means you can self-host the entire stack if needed, and the cloud offering is compatible with the self-hosted version; auto-generated REST and GraphQL APIs from schema eliminate boilerplate CRUD server code, letting teams focus on data modeling; Row Level Security integrates with Auth JWTs, so authorization is enforced by Postgres rather than scattered across app code; pgvector and 50+ extensions make it a viable backend for AI applications without adding a separate vector database.

Cons: Free projects pause after 1 week of inactivity and are limited to 2 active projects, making the free tier unsuitable for long-running low-traffic services; compute costs scale with instance size, and moving from shared CPU (Micro/Small/Medium) to dedicated CPU (Large+) is a significant price jump (~$10/mo to ~$110/mo); the API key migration from anon/service_role to publishable/secret keys requires code changes across clients and backends, and pg_net webhooks need special handling for the new key format; storage and egress costs add up beyond the included 8GB disk and 100GB cached egress; it is a backend platform, not a no-code tool—you still need to write SQL, configure RLS policies, and manage deployments, which is why AI builders like Lovable and Bolt use it as a backend rather than a replacement for their frontends.

Who is it for?:Developers building web, mobile, or AI applications who want a managed Postgres backend with auth, storage, realtime, and edge functions without stitching together separate services; teams that value open-source portability and want the option to self-host; AI application builders who need pgvector alongside relational data; developers using AI coding tools (Cursor, Claude Code) that can manage Supabase projects via MCP. Not suited for teams that need a fully no-code backend, projects that require a database to stay active without any traffic (Free tier pauses), or use cases where a simpler managed database without the full BaaS suite would suffice.