Supabase

Open-source backend-as-a-service built on Postgres + PGVector + auto-generated REST/GraphQL APIs + auth + storage + edge functions. Has become the default backend infrastructure referenced across this wiki — especially for agent memory systems, AI coding workflows, and visual agent builders. Self-hostable, with a generous free hosted tier.

  • Vendor: Supabase Inc.
  • Site: supabase.com
  • License: Open source (Apache 2.0 for the core, dual-licensed for some extensions)
  • Underlying tech: Postgres + PostgREST + GoTrue (auth) + PGVector (vector search) + Realtime (websockets)

Why It Keeps Showing Up

Across the ingested wiki content, Supabase is the convergence point for several distinct threads:

1. Agent Memory Infrastructure (OpenBrain)

Nate B Jones’s $0.10/month agent memory system is built on Supabase’s hosted Postgres + PGVector. The architectural argument: instead of paying for a proprietary memory product, run agent memory as a SQL database you own, exposed via MCP to any LLM. Supabase makes this practical because Postgres + PGVector + auth + REST API come bundled.

2. AI Coding Assistants Backend (ai-coding-workflow)

Cole Medin’s “Code 100x Faster with AI” video uses Supabase as the demo target — building a Supabase MCP server in a single Windsurf prompt. The workflow assumes Supabase is the database layer for the apps you’re building with AI coding assistants.

3. AI Coding Command Center (archon-os)

Cole’s Archon OS requires Supabase as its backing database. Setup involves creating a Supabase project, running a migration SQL file, and pointing Archon at the Supabase URL. Supabase is non-optional for Archon’s knowledge base + task management.

4. Visual Workflow Builder Integrations (sim-ai)

Sim AI ships with Supabase as one of its 60+ pre-built integration nodes — for both data storage and as a vector store for knowledge bases.

Why It Won (Per the Wiki Sources)

The pattern across all four uses is the same: a Postgres-shaped problem that also needs auth, vector search, and a REST API without writing one. Supabase replaces what would otherwise be:

Without SupabaseWith Supabase
Postgres install + maintenanceHosted Postgres
Write a REST API layerAuto-generated PostgREST
Pick + configure an auth libraryGoTrue auth bundled
Pick + configure a vector storePGVector extension included
Stand up a webhook systemRealtime built in
Pay for 5 separate SaaS toolsOne bill

For agent memory specifically, the killer feature is PGVector inside Postgres — vector similarity search against the same database that holds your structured agent state. No two-system sync problem.

Pricing

  • Free tier: 500 MB database, 1 GB file storage, 2 GB bandwidth, 50K monthly active users — enough for serious personal projects
  • Pro: ~$25/month — production tier, more compute, more storage
  • Enterprise: custom

The free tier is what makes OpenBrain’s “$0.10/month agent memory” claim possible — you’re effectively running on a free Supabase project until you outgrow it.

Self-Hosting

Supabase is fully open-source and can be self-hosted via Docker Compose. For privacy-sensitive setups (e.g., the offline knowledge server pattern) or for users who want complete data control, the self-hosted path is supported and documented. The trade-off is operational overhead vs the hosted convenience.

How It Compares

SupabaseFirebaseNeonConvex
DatabasePostgresFirestore (NoSQL)PostgresCustom reactive
Vector searchPGVector built inAdd-onPGVector add-onBuilt in
AuthBuilt inBuilt inAdd-onBuilt in
Auto REST APIYes (PostgREST)Yes (Firestore SDK)No (write your own)Yes (functions)
Self-hostableYesNoNoNo
Open sourceYesNoPartiallyNo
Free tierGenerousGenerousGenerousGenerous
Best forPostgres-shaped + open sourceMobile-firstPure Postgres + branchingReactive backend

See Also