This section is the product-native deep dive: how Weaviate's indexes, tenancy, and MCP capabilities underpin Engram's managed memory model — topics, scopes, pipelines, and retrieval types.
What you will learn
- Why Weaviate is a foundation for agent memory workloads
- Hybrid search, multi-tenancy, RBAC, and MCP as agent tool access
- Engram's units — memories, groups, topics, scopes, and bounded topics
- Engram pipelines, async runs, deduplication, and personalization templates
Chapters in this section
There are 24 chapters in this part. Open any chapter to read it on its own, or work through them in order.
- 1 Why Weaviate as a Foundation for Agent Memory
- 2 Weaviate’s Vector Index and Inverted Index Working Together
- 3 Native Hybrid Search in Weaviate
- 4 Multi-Tenancy as Weaviate’s Isolation Primitive
- 5 Role-Based Access Control in Weaviate
- 6 Collections and Schema Design in Weaviate for Memory Use Cases
- 7 Weaviate’s Model Context Protocol Server
- 8 Exposing Hybrid Search as an Agent Tool via MCP
- 9 Granular MCP Permissions for Memory Access
- 10 Weaviate Engram: A Managed Memory Service Built on Weaviate
- 11 Memories as Engram’s Core Unit
- 12 Groups: Bundling Topics and Pipelines in Engram
- 13 Topics: Configuring What Engram Extracts
- 14 Scopes in Engram: Project, User, and Property-Level Isolation
- 15 Bounded Topics and Single-Object-Per-Scope Guarantees
- 16 Engram’s Pipeline Architecture: Extract, Transform, Commit
- 17 Durable Workflow Execution Behind Engram Pipelines
- 18 Engram’s Input Data Types: Conversation, String, and Pre-Extracted
- 19 Retrieval Types in Engram: Vector, BM25, Hybrid, and Fetch
- 20 Asynchronous Processing and Run Status in Engram
- 21 Engram’s Approach to Deduplication and Reconciliation
- 22 Building Personalization Templates on Engram
- 23 Building Continual-Learning Templates on Engram
- 24 Integrating Engram Into Coding Assistants and Developer Tools