Summary
When using @frontmcp/sdk with the decorator-style API in TypeScript, autocomplete works for the first level of the @FrontMcp(...) config object, but nested objects do not provide proper autocomplete.
For example, I can get completion for top-level properties like auth, but once I enter auth: { ... }, TypeScript/IDE autocomplete is missing or incomplete.
Example
import 'reflect-metadata';
import { FrontMcp, LogLevel } from '@frontmcp/sdk';
import { CalcApp } from './calc.app';
@FrontMcp({
info: { name: 'Demo 🚀', version: '0.1.0' },
apps: [CalcApp],
logging: {
level: LogLevel.Debug,
},
auth: {
// expected nested autocomplete here, but it is missing/incomplete
},
})
export default class Server {}
Actual behavior
- Top-level autocomplete works for
@FrontMcp({...})
- Nested autocomplete inside objects like
auth is missing or much weaker than expected
Expected behavior
Nested config objects should preserve full IntelliSense / autocomplete, for example inside:
auth
http
pubsub
pagination
elicitation
- other nested config sections
Notes
This makes the decorator-style API harder to use, because the user can discover the top-level keys but not the inner structure.
It feels like the exposed decorator typing is not preserving nested IntelliSense well enough in editor usage.
Suggestion
One possible improvement would be to expose more explicit public interfaces/types for the config objects and derive schema validation from those types, rather than relying on schema-first inference only.
That may improve editor usability and make nested configuration autocomplete more predictable.
Summary
When using
@frontmcp/sdkwith the decorator-style API in TypeScript, autocomplete works for the first level of the@FrontMcp(...)config object, but nested objects do not provide proper autocomplete.For example, I can get completion for top-level properties like
auth, but once I enterauth: { ... }, TypeScript/IDE autocomplete is missing or incomplete.Example
Actual behavior
@FrontMcp({...})authis missing or much weaker than expectedExpected behavior
Nested config objects should preserve full IntelliSense / autocomplete, for example inside:
authhttppubsubpaginationelicitationNotes
This makes the decorator-style API harder to use, because the user can discover the top-level keys but not the inner structure.
It feels like the exposed decorator typing is not preserving nested IntelliSense well enough in editor usage.
Suggestion
One possible improvement would be to expose more explicit public interfaces/types for the config objects and derive schema validation from those types, rather than relying on schema-first inference only.
That may improve editor usability and make nested configuration autocomplete more predictable.