Inside Weaviate's MCP Ecosystem, Agent Memory with Engram and 4-bit Rotational Quantization

Hello Weaviate Community! 🤗

Curious how Weaviate's MCP ecosystem fits together? Our upcoming webinar takes a closer look, and we'd love to see you there. This edition also includes a practical guide to configuring Engram memory, the engineering behind 4-bit rotational quantization, and new podcast conversations about human behavior modeling and an AI-native insurance brokerage.

Latest AI & tech insights

Explore our recent Weaviate content:

Read

• ✍️ Agent Memory with Engram: A Practical Guide: Learn how topic descriptions control memory extraction, when to use bounded topics and scopes, how to select a retrieval mode, and where to place memories without hurting prompt-cache efficiency. Read the blog

• 🔖 4-bit Rotational Quantization: See how SIMD improvements accelerate rotations, encoding, and distance calculations in Weaviate 1.39. The post also examines scaling behavior, compares RQ with TurboQuant, and shows how 4-bit RQ provides similar recall with a 45% heap reduction. Read the blog

Watch

• 🎙 humans& with Alexis Ross, Manya Bansal, and Niloofar Mireshghallah - Weaviate Podcast #145: The humans& team introduces Persimmon, a research-preview user model for simulating human behavior in multi-turn, multi-party conversations. They discuss distribution matching, reward hacking, and a multi-turn Turing test where Persimmon reaches roughly 20% against a 50% ceiling. Watch the full podcast

• 🎙 Delegance Brokerage with Alex Ledbetter - Weaviate Podcast #144: Alex Ledbetter explains how he built an AI-native commercial insurance brokerage that processes binders approaching 1,000 pages, vectorizes documents in Weaviate, and uses three million carrier appetite rules to route clients. Watch the full podcast

Product highlights

Check out our latest product updates.

• 4-bit rotational quantization: Reduce vector heap usage by 45% while maintaining recall similar to 8-bit RQ.

• Faster SIMD operations: SIMD support speeds up Fast Walsh-Hadamard Transforms by up to 6.1x, while updated encoding and distance kernels also improve 8-bit RQ performance.

• 4-bit RQ is now available for HNSW collections on Weaviate Cloud.

Read the technical deep dive for the benchmarks and scaling analysis.

Upcoming Events

Join us online or in person for expert-led AI workshops and networking events.

• 🎫 Inside Weaviate's MCP Ecosystem | October 14: Join the webinar for a closer look at Weaviate's MCP ecosystem. In this webinar, we’ll explore why MCP matters and demonstrate two MCP servers in the Weaviate ecosystem: Weaviate Database and Weaviate Query Agent MCP. Register here

Hungry for more?

Have a question or want to connect? Join our Weaviate Forum to engage in community conversations.

Follow our Blog, LinkedIn, Twitter, and YouTube for more content updates.

See you next month,

Shan