Memory Search Server

By MEOK AI Labs — Persistent AI memory system with semantic search, care-weighted episodes, and knowledge base

Installation

pip install memory-search-mcp

Usage

python server.py

Tools

record_memory

Record a memory episode with care-weighting and emotional valence. Memory types: interaction, insight, decision, emotion.

Parameters:

• content (str): Memory content
• source_agent (str): Source agent (default: "user")
• memory_type (str): Type: interaction, insight, decision, emotion (default: "interaction")
• care_weight (float): Retrieval priority 0-1 (default: 0.5)
• importance (float): Importance 0-1 (default: 0.5)
• emotional_valence (float): Emotional valence 0-1 (default: 0.5)
• tags (list[str]): Tags for categorization
• parent_id (str): Parent episode ID for chains

search_memory

Semantic search across all memories using full-text search with relevance ranking.

Parameters:

• query (str): Search query
• limit (int): Max results (default: 10)
• care_weight_min (float): Minimum care weight filter (default: 0.0)
• memory_type (str): Filter by type
• tags (list[str]): Filter by tags

add_knowledge

Add a knowledge entry to the persistent knowledge base (facts, definitions, reference material).

Parameters:

• topic (str): Knowledge topic
• content (str): Knowledge content
• source (str): Source (default: "manual")
• confidence (float): Confidence level (default: 0.8)
• tags (list[str]): Tags

search_knowledge

Search the knowledge base by topic or content.

Parameters:

• query (str): Search query
• limit (int): Max results (default: 10)
• min_confidence (float): Minimum confidence (default: 0.0)

list_memories

List recent memories, optionally filtered by type. Sort by created_at, care_weight, importance, or access_count.

get_memory_stats

Get statistics about the memory store: total count, type breakdown, average care weight, most accessed, storage size.

get_temporal_chain

Follow the temporal chain from a memory episode forward or backward in time.

consolidate_memories

Consolidate old, low-access memories by summarizing and archiving them.

semantic_search

Semantic search using TF-IDF cosine similarity (no external dependencies).

Authentication

Free tier: 100 calls/day. Upgrade at meok.ai/pricing for unlimited access.

License

MIT — MEOK AI Labs