Convex

Convex is a backend platform rather than a database on its own. It provides a document store, server-side functions written in TypeScript that run inside the platform, scheduled jobs and workflows, file storage, text and vector search, and a reactive query layer. Queries are deterministic functions over the database; Convex records what each one read and re-runs it for subscribed clients when any of that data changes, so a React or mobile client receives updated results without polling or manually invalidating a cache. Mutations are transactions, so a function either commits entirely or not at all. The backend source is published under FSL-1.1-Apache-2.0, the Functional Source Licence, which permits self-hosting and internal use, forbids competing use, and converts each release to Apache 2.0 two years after it was made available. The thing that distinguishes it is that reactivity is a property of the database, not a library on top of one. Building the same behaviour on Postgres means change data capture, a subscription server, cache invalidation logic and a plan for what happens when the two drift; Convex makes that the default path and the correctness of it the vendor's problem. The commercial consequence is that this is not a component you can swap later. Your data access, your authorisation checks, your transaction boundaries and your client subscription model are all written against Convex's function API, so the application is coupled to the platform at a much deeper level than a connection string. The buyers are product teams building collaborative or live-updating applications, and increasingly teams building agent backends, who would otherwise spend months on infrastructure that is not their product. The trade-off is the transaction envelope. A query or mutation gets one second of user code, may read and write at most 16 MiB, may scan 32,000 documents and write 16,000, and each document is capped at 1 MiB, so any bulk operation must be chunked and scheduled rather than run as a single transaction. There is no SQL and no query planner; you declare indexes, up to 32 per table, and traverse them, which means an access pattern you did not anticipate is a schema change rather than a new query. That is a reasonable trade for a young product moving fast and a poor one for a system whose query patterns are unknown.

• 

  Convexvs

  

  Cockroach Labs

  Cockroach Labs: The cloud-native distributed SQL database

• 

  Convexvs

  

  PostgreSQL

  PostgreSQL: The world's most advanced open source relational database

• 

  Convexvs

  

  Airtable

  Airtable: Create apps that perfectly fit your team's needs

• 

  Convexvs

  

  Amazon Aurora

  Amazon Aurora: MySQL and PostgreSQL-compatible relational database built for the cloud