Search Persistent Agent Memory

CLI Tool Name: memory_search

Full-text searches across all memory: facts, strategy notes, and checkpoint metadata. This avoids explicitly recalling keys utilizing memory_read and enables broad sweeping index access for your agent locally.

Parameters

Parameter	Type	Required	Description
query	string	yes	Case-insensitive keyword or phrase to search across all memory sources.

Example output

json

{
  "results": [
    {
      "key": "user_preferred_style",
      "value": "Tailwind CSS",
      "category": "preferences"
    },
    {
      "key": "css_framework_decision",
      "value": "We chose Tailwind over standard CSS to retain velocity.",
      "category": "strategy"
    }
  ],
  "count": 2
}

How to use it

Your AI client calls this tool automatically when searching for lost context or auditing its own memory:

example prompt

Search my memory for any strategy notes or facts related to 'CSS'.

[tip]

If you know the exact key for the context you need loaded, use memory_read instead as it's faster and saves tokens.

Explore other AgenticStore MCP tools

→ Memory Write MCP Tool → Memory Read MCP Tool

← memory_read overview →