SDKs as a Convenience Layer Over REST

The SDK Wraps the REST API

Claude is most commonly consumed through the official Python and TypeScript SDKs. These SDKs are a convenience layer, not a separate API surface -- underneath them are ordinary HTTPS + JSON calls to the same REST API. The SDK handles:

• Authentication (API key management)
• Request/response serialization (turning SDK objects into JSON and back)
• Typed errors (structured exceptions instead of raw HTTP status codes)
• Retries (automatic backoff/retry on transient failures)
• Streaming helpers (consuming a server-sent event stream as an iterator)

You Can Always Drop to Raw REST

Because the SDK is a convenience layer and not a different protocol, a developer can always bypass it and make the equivalent raw HTTPS + JSON request directly -- useful in environments without an official SDK, or when fine-grained control over the request is needed.

Sync vs. Async Clients

Both SDKs typically offer a synchronous client (each call blocks until it completes) and an asynchronous client (calls can be issued concurrently). Use the async client to parallelize independent, I/O-bound calls -- for example, issuing several unrelated model calls concurrently rather than awaiting each one in sequence -- which reduces total wall-clock time without changing token cost.

Common exam traps

• Treating the SDK as a fundamentally different API from REST, rather than a typed convenience wrapper over the same HTTPS + JSON calls.
• Assuming a sync client can't be parallelized at all, or that switching to async changes billing -- async changes concurrency/wall-clock time, not token cost.