Embeddings, agents and APIs on one Postgres.
Store embeddings next to your data with pgvector, queue embedding jobs with pgmq, and let coding agents build the backend over MCP.

pgvector built in
Store and search embeddings with pgvector and HNSW indexes, filtered by RLS.
Background jobs
Queue embedding and model jobs in Postgres and process them from your server, with retries and visibility timeouts.
Agent-ready
Every project has an MCP server so Claude Code or Cursor can design schemas and write migrations.
Realtime streaming
Stream tokens and progress to clients over Realtime Broadcast.
Dedicated compute
Vector search is CPU-hungry. Pro and Business run on dedicated VMs with 8 and 16 GB of RAM.
Your data stays yours
Embeddings live in your own Postgres, exportable at any time.
Questions
Which vector index types are supported?
pgvector's IVFFlat and HNSW indexes are both available.
Can I run models on SupaHost?
SupaHost hosts your backend, not models. Call hosted model APIs from your server, or from Postgres with pg_net.
Put every project on one plan.
Every plan starts with a 7-day free trial.