Skip to content

search_memories

Search memories by semantic similarity to a natural language query.

Parameters

ParameterTypeRequiredDescription
querystringYesA natural language description of what you're looking for.
top_kintegerNoMaximum number of results (1–50, default 10).

Behaviour

  • Converts your query to a vector embedding and finds the most semantically similar memories.
  • Results are ranked by relevance (score from 0.0 to 1.0, higher = more relevant).
  • Returns up to top_k results — there is no pagination.
  • Returns an empty list if no memories have been stored yet.

Examples

Search my memories for anything about authentication.
Find memories related to the database schema.
What have I stored about project deadlines?

How it differs from recall and list_memories

ToolUse when
recallYou know the exact key
list_memoriesYou want everything with a specific tag
search_memoriesYou remember roughly what it was about but not the key or tag

Hive — shared persistent memory for AI agents