Redesign Claude SDK source as a governed Kohala agent
Have Kai statically inspect Anthropic Messages SDK or Claude Agent SDK source and propose a native Kohala agent design for approval. This design-and-migration path does not promise unchanged provider-model execution.
Explore Kohala
Frequently Asked Questions
- Can Kohala work from my Claude SDK source?
- Yes. Upload Anthropic Messages SDK or Claude Agent SDK source to Kai. The Claude Engineer statically inspects the scoped files, explicitly distinguishes the two SDKs, and proposes a native Kohala design with dependencies, credential names, migration mapping, and fidelity limits. It never executes the upload during inspection.
- Will Kohala run my Claude provider loop unchanged?
- No claim is made that it will. This is a design-and-migration path, not unchanged native provider-model execution. You review the proposed native Kohala scripts and capabilities, and nothing is built until you approve through the normal lifecycle gate.
- Are Claude API keys copied into the design?
- No. The specialist records referenced credential names and dependency requirements, not credential values. Secrets are configured separately through Kohala's account-scoped secret controls.