Display Name
mcp-memory-service
Category
Tooling
Sub-Category
General
Primary Link
https://github.com/doobidoo/mcp-memory-service
Author Name
doobidoo
Author Link
https://github.com/doobidoo
License
Apache-2.0
Other License
No response
Description
MCP-native persistent memory service for Claude Code with semantic search, knowledge graph, entity extraction, insight cards, and autonomous memory consolidation. Provides 25+ MCP tools including memory_store, memory_search (hybrid BM25+vector), memory_harvest (session learning), mistake notes, and quality scoring. Supports sqlite-vec (local), Milvus, and Cloudflare backends. REST API with OAuth 2.1 for multi-user deployments. Used in production with 9,300+ stars.
Validate Claims
Install via npx: npx -y @doobidoo/mcp-memory-service
Or add to claude_desktop_config.json as an MCP server. The service starts immediately with zero configuration (sqlite-vec backend, local embeddings). No API keys needed for basic usage.
Specific Task(s)
Store a few memories about your project decisions, then ask Claude to recall them. Enable entity extraction (MCP_ENTITY_EXTRACTION_ENABLED=true) and run memory_quality(action="maintain") to see insight cards generated from your memory corpus.
Specific Prompt(s)
"Store this decision: we chose FastAPI over Flask because of async support and automatic OpenAPI docs." Then later: "What framework decisions have we made?" — Claude will use memory_search to retrieve the stored decision with semantic matching.
Additional Comments
Active contributor here (tag_match filtering, session chunking, locale-based harvest patterns, plugin system). The v10.55-v10.59 arc added entity extraction, insight cards, temporal contradiction detection, and Kiro CLI harvest support — transforming it from a storage layer into an intelligence tier. Weekly releases, responsive maintainer.
Recommendation Checklist
Display Name
mcp-memory-service
Category
Tooling
Sub-Category
General
Primary Link
https://github.com/doobidoo/mcp-memory-service
Author Name
doobidoo
Author Link
https://github.com/doobidoo
License
Apache-2.0
Other License
No response
Description
MCP-native persistent memory service for Claude Code with semantic search, knowledge graph, entity extraction, insight cards, and autonomous memory consolidation. Provides 25+ MCP tools including memory_store, memory_search (hybrid BM25+vector), memory_harvest (session learning), mistake notes, and quality scoring. Supports sqlite-vec (local), Milvus, and Cloudflare backends. REST API with OAuth 2.1 for multi-user deployments. Used in production with 9,300+ stars.
Validate Claims
Install via npx: npx -y @doobidoo/mcp-memory-service
Or add to claude_desktop_config.json as an MCP server. The service starts immediately with zero configuration (sqlite-vec backend, local embeddings). No API keys needed for basic usage.
Specific Task(s)
Store a few memories about your project decisions, then ask Claude to recall them. Enable entity extraction (MCP_ENTITY_EXTRACTION_ENABLED=true) and run memory_quality(action="maintain") to see insight cards generated from your memory corpus.
Specific Prompt(s)
"Store this decision: we chose FastAPI over Flask because of async support and automatic OpenAPI docs." Then later: "What framework decisions have we made?" — Claude will use memory_search to retrieve the stored decision with semantic matching.
Additional Comments
Active contributor here (tag_match filtering, session chunking, locale-based harvest patterns, plugin system). The v10.55-v10.59 arc added entity extraction, insight cards, temporal contradiction detection, and Kiro CLI harvest support — transforming it from a storage layer into an intelligence tier. Weekly releases, responsive maintainer.
Recommendation Checklist