Both give you an instant backend over a real database. The difference is where it runs, how you model collections, and how AI agents reach your data.
Supabase is an excellent open-source backend built around a single managed PostgreSQL database, with PostgREST, Realtime, Auth, and Storage on top. backlex targets a different shape: one codebase that deploys to Cloudflare Workers, Vercel, Netlify, or your own server, runs on either PostgreSQL or SQLite/D1, and ships a first-class runtime MCP server so AI agents can use your backend as a tool.
| backlex | Supabase | |
|---|---|---|
| License | Apache-2.0 core + managed Cloud | Apache-2.0 + managed cloud |
| Database | PostgreSQL and SQLite / Cloudflare D1 (dual-dialect) | PostgreSQL only |
| Runtimes | Cloudflare Workers, Vercel & Netlify (Node 22), self-host on Bun — one codebase | Managed cloud; self-host via Docker |
| Edge-native | Runs inside the edge runtime, next to the request | Regional Postgres + Deno edge functions in front |
| Collections / schema | Managed tables (dynamic, additive-only DDL) or adopt existing tables via a metadata row — no data migration | Manage Postgres schema directly (SQL / migrations / Studio) |
| Field types | text, longtext, number, boolean, uuid, timestamp, json, relation / relation_many, i18n_text | Native Postgres column types |
| REST API | Auto-generated; filter/sort/paginate, nested-relation filters (2 hops), expand, aggregation | Auto-generated via PostgREST |
| GraphQL | Built in, auto-generated from collections | Via pg_graphql extension |
| Realtime | Live row changes streamed to clients over Server-Sent Events (SSE), permission-filtered | Postgres logical replication over WebSockets |
| Permissions | Role-based DSL: conditions + field-level allow-lists, one model across REST/GraphQL/realtime | Postgres Row-Level Security (SQL policies) |
| Vector search | Built in — pgvector (PG) or Cloudflare Vectorize (SQLite/D1); auto-embed on write | Built in (pgvector) |
| Embedding models | OpenAI, Workers AI (bge-m3), or self-hosted TEI/Ollama/vLLM | Bring your own (OpenAI, etc.) |
| AI agents / MCP | Built-in runtime MCP server — collections, vector search & admin exposed to agents, per-key tool allow-lists + read-only guards | Official MCP server (project / database management from dev tools) |
| Storage + images | Files with ACLs, signed URLs; image transforms (resize/format/fit) on Bun in-process or the Cloudflare edge | Storage with image transformations |
| Edge functions | Sandboxed functions (QuickJS / worker-thread / remote executor); http/event/cron triggers | Edge Functions (Deno) |
| Auth providers | Email/password, OAuth, magic link, email-OTP, passkeys, anonymous | Email/password, OAuth, magic link, OTP |
| SSO | SAML 2.0 (per-workspace) and LDAP/AD | SAML (Pro / Enterprise) |
| Audit logs | Built-in activity trail + opt-in sensitive-read auditing | Via Postgres / log drains |
| Flows (automation) | Visual, no-code op-chains triggered by item events | Database triggers / functions (no visual builder) |
| Webhooks | Signed webhooks on item CRUD events | Database webhooks |
| Admin UI + SDK | Admin SPA, typed client SDK, and `backlex` CLI | Studio, supabase-js, CLI |
In backlex you either let it create a managed table (dynamic schema with an additive-only DDL applier that never drops or rewrites columns) or adopt an existing table by writing a single metadata row — no data migration, foreign keys auto-detected and surfaced as relations. Every collection instantly gets permission-aware REST + GraphQL, realtime, and an admin UI. With Supabase you manage the Postgres schema yourself (SQL, migrations, or the Studio table editor) and wire RLS policies per table.
Mark a collection vectorizable and flag the text fields to embed —
backlex auto-embeds on every
write and runs approximate-nearest-neighbour search, using
pgvector on Postgres
or Cloudflare Vectorize
on SQLite/D1. Embeddings come from OpenAI, Workers AI (bge-m3), or your
own self-hosted model. The same vectors back the Ask-AI / RAG page and
the vector.search MCP tool, so agents get semantic search
for free. Supabase also offers vector search via pgvector, but you wire
the embedding pipeline and any agent access yourself.
See the backlex documentation or the source on GitHub to get started. Self-hosting is free under Apache-2.0; the managed Cloud is optional.