Store NEW information. Use for preferences, decisions, facts.

ACTION GUIDE — when to use:

• Use when saving new information for the first time. Example: content='User prefers dark mode', category='preference', tags=['ui']

Find existing memories by natural language query. Always search before adding.

ACTION GUIDE — when to use:

• Use BEFORE adding new information to avoid duplicates. Example: query='dark mode preference'

Browse all memories, optionally filtered by category.

ACTION GUIDE — when to use:

• Use when you want to view a broad set of memories, or see what's in a specific category. Example: category='preference', limit=10

Modify an EXISTING memory by ID. Get memory_id from search results.

ACTION GUIDE — when to use:

• Use when an existing fact or preference changes. Example: memory_id='abc123', content='User now prefers light mode'

Remove a memory by ID.

ACTION GUIDE — when to use:

• Use when a memory is completely outdated, incorrect, or explicitly requested to be forgotten. Example: memory_id='abc123'

Export all memories as JSONL.

ACTION GUIDE — when to use:

• Use when the user requests a backup or raw dump of their memory database.

Import memories from JSONL data or a list of objects.

ACTION GUIDE — when to use:

• Use when restoring from a backup or migrating data into the memory system. Example: data='[{"content": "example"}]', mode='merge'

Show database statistics (total memories, categories, embedding status).

ACTION GUIDE — when to use:

• Use when you need to understand the scale or health of the memory database, or check if embeddings are enabled.

Restore an archived memory by ID.

ACTION GUIDE — when to use:

• Use to bring a previously archived memory back into the active search pool. Example: memory_id='abc123'

List archived memories.

ACTION GUIDE — when to use:

• Use to view memories that have been soft-deleted or automatically archived due to low importance/recency.

Summarize similar memories in a category (requires LLM API keys).

ACTION GUIDE — when to use:

• Use when a category has too many redundant or closely related memories and needs cleanup. Example: category='preference'

[DEPRECATED — use the granular tools (add_memory, search_memory, ...) instead; this composite tool will be removed in a future release]

Legacy dispatcher for backward compatibility. Use specialized tools (add_memory, search_memory, etc.) instead.

Persistent memory store. Actions: add|search|list|update|delete|export|import|stats|restore|archived|consolidate.

ACTION GUIDE — when to use each:

• add: Store NEW information. Requires 'content'. Use when saving preferences, decisions, facts for the first time. Example: action='add', content='User prefers dark mode', category='preference', tags=['ui']

• search: Find existing memories by natural language query. Requires 'query'. Use BEFORE add to avoid duplicates. Example: action='search', query='dark mode preference'

• update: Modify an EXISTING memory by ID. Requires 'memory_id' (from search/list results). Use when a fact changes. Example: action='update', memory_id='abc123', content='User now prefers light mode'

• list: Browse all memories, optionally filtered by category. No query needed.

• delete: Remove a memory by ID. Requires 'memory_id'.

• stats: Show database statistics (total memories, categories, embedding status).

• export: Export all memories to JSONL format.

• import: Import memories from JSONL data. Requires 'data'.

• archived: List archived memories. Optionally filter by limit.

• restore: Restore an archived memory by ID. Requires 'memory_id'.

• consolidate: Summarize and consolidate similar memories in a category using LLM. Requires 'category'.

WORKFLOW: search -> not found? -> add. Found outdated? -> update (with memory_id from results). PROACTIVE: save user preferences, decisions, corrections, project conventions.

Server config, sync, setup, and bounded vector backfill. Actions: status|sync|set|warmup|setup_sync|backfill_embeddings.

ACTION GUIDE — when to use each:

• status: Show current configuration, setup status, and database stats.

• sync: Trigger manual sync (requires sync_enabled=true + google_drive_client_id).

• set: Update a setting. Requires 'key' and 'value'. Valid keys: 'sync_enabled' (true/false), 'sync_interval' (int), 'log_level' (str). Example: action='set', key='sync_enabled', value='true'

• warmup: Pre-download embedding model (~570 MB) to avoid delays later.

• backfill_embeddings: Embed active rows missing a current Vectorize ledger entry; optional batch_size is bounded to 100.

• setup_sync: Authenticate Google Drive via Device Code OAuth flow.

Full documentation for memory and config tools. topic: 'memory' | 'config'

ACTION GUIDE — when to use:

• Use when you need detailed instructions on how to use specific server tools or features.

Get the relay configuration URL for mnemo-mcp, opening it in the user's browser when possible. Returns the relay URL, whether the browser launched, and the current status.