I get the "Why Claude Code", just not the "Why a REST API". It doesn't seem to solve a problem at all.
Check the "Who It's For"
> Solo founders who use Claude Code daily and want to use it in toolchains
I can do that just fine without a REST API
> Backend developers adding AI capabilities to services and APIs
You can interact with Claude Code just fine. In fact I wouldn't use CC here since I want control over how it's ran, and CLAUDE.md just doesn't have that power.
> AI tinkerers building agentic workflows, automation, and experiments
Ditto.
> Data engineers who need structured, typed LLM output in pipelines
This has nothing to do with the REST part. I don't see how this is better than just an SDK with structured output.
> Platform teams exposing Claude Code to internal services
I guess I see the appeal a bit here, but anthropic has a platform that can handle agents for you, and this seems like a better fit than a clunky REST API between a CLI tool.
Unless you're creating a disposable throwaway filesystem that's isolated between API calls, you're setting yourself up for a whole lotta stateful shenanigans...
I get the "Why Claude Code", just not the "Why a REST API". It doesn't seem to solve a problem at all.
Check the "Who It's For"
> Solo founders who use Claude Code daily and want to use it in toolchains
I can do that just fine without a REST API
> Backend developers adding AI capabilities to services and APIs
You can interact with Claude Code just fine. In fact I wouldn't use CC here since I want control over how it's ran, and CLAUDE.md just doesn't have that power.
> AI tinkerers building agentic workflows, automation, and experiments
Ditto.
> Data engineers who need structured, typed LLM output in pipelines
This has nothing to do with the REST part. I don't see how this is better than just an SDK with structured output.
> Platform teams exposing Claude Code to internal services
I guess I see the appeal a bit here, but anthropic has a platform that can handle agents for you, and this seems like a better fit than a clunky REST API between a CLI tool.
Unless you're creating a disposable throwaway filesystem that's isolated between API calls, you're setting yourself up for a whole lotta stateful shenanigans...
> With a typical LLM SDK
What is a typical LLM SDK?