Questions you are probably asking
Are these official Convex clients?
No. They are educational experiments built for this project, the website and the accompanying video. They are not official Convex SDKs, not published packages, and they will not be maintained or supported going forward. For real work, use the official Convex client libraries.
How much of the Convex API do they cover?
The core client surface: queries, mutations, actions, JSON-safe values, auth-token forwarding and error handling, plus realtime query subscriptions over WebSockets for the Live badge. They do not cover the full official SDK surface.
Is this server-side Convex?
No, client-side only. The backend the clients talk to is a normal Convex backend with its queries, mutations and actions written in TypeScript as usual. This project is about calling into a Convex backend from each language, the way a frontend or another service would. So no, sadly there is no COBOL running on Convex's servers.
Did AI really write all of it?
Yes, all of it. The human job was direction: picking the goals, setting the rules, reviewing the evidence and deciding what happened next. The agents did the typing, and the test suite kept everyone honest.
How do I know the badges are real?
Every badge comes from ./run verify-all at an exact
clean commit: 31 of 31 conformance checks against a local backend
and 31 of 31 against a real hosted deployment over TLS. The evidence
behind every badge, including the commit, container digest,
toolchain and timestamp, is on this site for every language.
How do I try one myself?
Install Docker, clone the
repository,
and run ./run verify-example <language> from the
repository root. That builds the language's pinned toolchain in
Docker and runs its canonical example against a real backend, so
nothing needs to be installed on your machine except Docker itself.
How is the roster still 100 after all those replacements?
When a language could not meet the shared rules, another viable candidate took its place. The graveyard is the final record of all 31 languages outside the finished roster: 27 ruled out by project constraints, three replaced during final host revalidation, and one passing candidate that arrived after the last slot was filled.
Can I get a proper idiomatic client for language X?
Maybe! If enough people want a fuller implementation of a particular language, it could happen. Say so in the video comments.