Postgres development platform
Supabase
Supabase provides an open-source development platform built around Postgres, giving developers the backend services needed to build applications. It combines databases, authentication, file storage, Edge Functions, realtime synchronization and automatically generated REST APIs. Developers can store, index and search vector embeddings for AI applications, secure data with row-level permissions, and write custom server-side code without managing servers. Its components work together or individually, and developers can inspect, contribute to and self-host the platform.
Founded in 2020 by CEO Paul Copplestone and CTO Ant Wilson, Supabase gives each project an isolated Postgres cluster so customers can scale independently. Its engineering work extends into database infrastructure: Multigres is a Postgres proxy designed to provide connection pooling, high availability and eventually sharding while prioritizing compatibility with the Postgres ecosystem. Its June 2026 release was a self-hostable alpha, with horizontal scaling still planned. Complementary work on OrioleDB develops a scalable storage engine for Postgres.
In June 2026, Supabase reported nearly 10 million developers, more than double its count eight months earlier, and 600% growth in database launches over the preceding year. More than 60% of new databases were launched through AI tools. That month, it raised a $500 million Series F led by GIC at a $10 billion pre-money valuation.
3 talks
Newest first
Skill Issue: How We Used AI to Make Agents Actually Good at Supabase
AI Engineer Europe 2026 · 1:18:41
2 speakers at AIE
Affiliations reflect their AIE appearances, not necessarily current employment.
Start here
- Combine Skills and MCP to Close the Context Gap
Start here to understand why access to MCP tools still leaves agents needing product-specific guidance and opinionated development workflows.
Pedro RodriguesAI Engineer Europe 2026
- Skill Issue: How We Used AI to Make Agents Actually Good at Supabase
Use the workshop to learn how SKILL.md and supporting scripts fit into building and refining agent skills.
Pedro RodriguesAI Engineer Europe 2026
- Supabase Vector: The Postgres Vector database
For infrastructure context, learn how pgvector fits alongside authentication, edge functions, and realtime features in embedding-powered applications.
Paul CopplestoneAI Engineer Summit 2023
Messages from the stage
Database context affects tool safety
Rodrigues illustrates how an agent-created PostgreSQL view can bypass row-level security unless security_invoker = true is set. His reported evaluations found MCP plus skills outperformed baseline and MCP-only conditions.
Evaluate changes to agent skills
Rodrigues’s workshop emphasizes evaluation scenarios, outcome inspection, and LLM-as-a-judge techniques to test whether skill revisions improve performance on Supabase workflows.
Vector indexing involves measurable tradeoffs
Copplestone compares IVFFlat and HNSW through throughput and accuracy, then discusses PostgreSQL partitioning and potential Citus work for applications storing billions of embeddings.
Affiliations reflect each recorded session, not necessarily current employment.