You
Which protocol for service to service calls?
ChatGPT
gRPC when the callers are yours and latency matters, REST when they are not:
| REST/JSON | gRPC | |
|---|---|---|
| Debuggable with curl | yes | awkward |
| Schema enforced | optional | always |
| Streaming | clumsy | native |
| Browser support | direct | needs a proxy |
The schema is the real difference: with JSON the contract lives in documentation, with protobuf it lives in a file that breaks the build when violated.