Short answer: A request-scoped assistant is useful for editing and running the call in front of you; a spec-driven assistant reasons over the whole contract and can generate tests, mocks, and external tools that are valid by construction. The first acts on a request; the second acts on the API.

What a request-scoped assistant sees

The active request, its response, and the environment. That is enough to tweak a header, adjust a body, and re-run the call, and it is genuinely handy for the current task.

What a spec-driven assistant sees

ScopeRequest assistantSpec-driven
ContextActive requestWhole contract
Can generateEdits to the callTests, mocks, docs, tools
ValidityAs editedBy construction
External agentsNot exposedMCP server

Why the scope matters

Generating a test or a mock from a single request risks inventing the rest of the API. Generating from the contract uses the declared operations and schemas, so the output is correct across the whole surface.

How they fit

Use the request assistant for the quick edit and the spec-driven assistant for anything that spans the API. The boundary follows the size of the task.

This is part of the Powerduck vs Hoppscotch comparison. Next, read testing in CI.