search-memory

Parallel memory search uses subagents to scan categorized memory files simultaneously, retrieving and aggregating historical project context into a single ranked answer. It targets broad recall queries spanning multiple memory files.

Cross-memory discovery is needed for broad conceptual questions like "what do we know about X" that could live in multiple files. Direct memory lookup is better for simple, known-file searches that can skip the parallel workflow.

Parallel subagent search spawns four agents simultaneously, mapping memory categories like Knowledge and Project State to distinct files, then aggregates and ranks their findings into one answer.

Yes, retrieving information from past errors and live context is possible with a single query. The parallel search spans all categorized memory files, including Errors & Guards and Live Context, to surface relevant data.

The best way to check if a problem was solved before is triggering cross-memory search. Asking "have we solved X before" executes parallel subagent searches across all memory categories to find historical solutions.

Cross-memory search handles both scenarios. While it uses multi-agent orchestration for broad recall across multiple files, simple single-file lookups can bypass the parallel workflow and execute directly.