MCP Completion

MCP completion is about helping clients suggest valid values while a user is filling prompt or resource-template arguments. It is closer to autocomplete than to an LLM generating an entire answer.

Why it matters

If a prompt accepts a repository name or a resource template expects a project identifier, completion can make the UI more usable and reduce invalid values.

Do not confuse with model completion

“Completion” in this context is not the same as asking an LLM to generate text. It is protocol support for value suggestions around MCP-exposed interfaces.

Design guidance

Keep completion fast, bounded, and permission-aware. Suggestions should never reveal resources the current caller cannot access.