Supabase
Supabase is an open-source backend platform launched in 2020 that wraps a managed PostgreSQL database with auth, object storage, realtime subscriptions, serverless edge functions (Deno) and vector search via the Postgres extension pgvector.
Supabase is an open-source backend platform built on PostgreSQL: database, authentication, storage, realtime and edge functions in one package. For agents it is usually the data layer, including vector search via pgvector.
Explanation
At its core it is deliberately ordinary Postgres: tables, SQL, row level security. Around it sit services you would otherwise assemble yourself — user management, file storage, realtime channels, functions at the network edge. For agentic applications this is attractive because an agent needs three things: a place for knowledge (embeddings in pgvector), a place for state (conversations, jobs, approvals) and a protected interface through which it calls tools (edge functions). Row level security is not a detail here but the actual security anchor: it decides which rows a user — and therefore an agent acting on their behalf — may see at all. Because everything is open source, the same architecture can be self-hosted, which simplifies data-sovereignty discussions considerably.
Marketing Relevance
For marketing and technology teams the data layer determines how quickly a prototype becomes an operable product. Supabase shortens that path because lead data, content, knowledge base and approval processes all live in a familiar SQL world — auditable, exportable and without a special-purpose vector store. For GDPR projects it also matters that an EU region can be selected and a data processing agreement signed.
Example
A knowledge base for the website chat lives as a table with a pgvector column in Supabase. An edge function takes the visitor question, retrieves the matching passages, calls the language model and writes question, answer and sources back as a conversation log. Row level security ensures only the internal team can read those logs.
Common Pitfalls
The most common mistake is tables without effective row level security policies — a public read key in the frontend then amounts to an open database. Just as often teams overlook that every project has one primary region: if you promise data sovereignty, choose that region deliberately instead of accepting the default.
Origin & History
Supabase launched in 2020 as an open-source answer to Firebase and bet on Postgres rather than a bespoke database from day one. With the rise of RAG applications from 2023, pgvector became the standard way to store embeddings without a separate vector database. By 2026 the platform is available in many regions, several of them in the EU.
Further Resources
Marketing Use Cases
Knowledge base with embeddings for a website or support chat.
Lead capture from forms including logging and notification.
State store for agents: jobs, intermediate results, approvals.
Tool endpoints for agents as edge functions with permission checks.
Frequently Asked Questions
Do I need a separate vector database alongside Supabase?
In most marketing applications, no. pgvector handles knowledge bases well into the range of tens of thousands of passages. A specialised vector database only pays off when volume, filter logic or query frequency go clearly beyond that.
Can Supabase be operated in a GDPR-compliant way?
Yes, if three things are in place: an EU region, a signed data processing agreement and a documented deletion concept for personal data. The platform provides the foundation; responsibility for purposes and retention stays with the company.
How does Supabase differ from Firebase?
Supabase builds on relational Postgres with SQL and open standards, Firebase on its own document-oriented database. The practical difference shows up in analysis and in switching: SQL data is easier to evaluate and to migrate.
Related Services
Go deeper: Agentic AI Hub · Governance & compliance