Never let a signup go cold again.
Turn emails into revenue with Brew. No credit card, free credits to try.
Turn emails into revenue with Brew. No credit card, free credits to try.
materialize.com · newsletter
Explore this email design and adapt it to your own brand. Review the copy, links, and offer before sending.
View in browser
Materialize
NEWSLETTER JUNE 2026
Building the live context graph for agents
As AI initiatives move from pilot to production, many engineering teams are finding that the bottleneck is fast, reliable access to fresh business context. AI agents in particular run on a tight feedback loop of observing the world, taking an action, and seeing the result, so it's vital that they're acting on data they can trust. As a result, more and more teams are optimizing for 'time to trusted action,' the time between something changing in the business and an agent being able to act on it with confidence, as Seth, our Field CTO, walks through in a recent webinar.
But minimizing that time is hard, as traditional architectures all have failure modes. Powering agents directly from siloed operational databases often means slow, complex queries that can't return fast enough and can degrade performance. Teams using vector databases or search indexes often face a cost-versus-freshness trade-off, where typical pipelines either update embeddings and attributes too often, driving costs up, or not often enough, meaning stale data. And custom streaming pipelines come with their own challenges: they're fragile, expensive to change and operate, and it’s difficult to ensure correct, consistent results that you can trust.
What these teams are converging on instead is a new architectural pattern, the 'live context graph': an interconnected set of data products representing core business entities, like Customers, Orders, or Shipments. Materialize helps teams build this graph just using SQL, and maintains data products incrementally as the underlying data changes, so agents can read from a single, real-time layer that's always fresh. That keeps time to trusted action low, and teams can spend less time wiring pipelines and more time building features that matter to the business.
Many companies are running this pattern with Materialize in production today, in industries ranging from SaaS to financial services to logistics. And we're continuing to make Materialize even faster, simpler, and easier for running this pattern, with our latest blog post covering everything we've shipped in the first half of this year. Want to see it in action? Schedule a demo with our team, and we can work through your use case.
-The Materialize Team
🚀 What's New in Materialize
Since our last newsletter, we've shipped new MCP servers for agents and developers, a new CLI for declarative deployments, and Single Sign-On for Self-Managed.
MCP Server for Agents. We now have a built-in MCP server for agents in public preview, so AI agents can discover and query the data products you've published in Materialize through /api/mcp/agent. This makes it easy to wire agents into Materialize through a single endpoint, with role-based access control over which agent can see what.
MCP Server for Developers. Also in public preview, our MCP server for developers gives coding agents diagnostic access to your Materialize environment. Point your coding agent at /api/mcp/developer and ask natural-language questions, like "why isn't this view fresh?", to troubleshoot freshness, hydration, and resource issues without leaving your editor. It pairs with our coding agent skills, which give Claude Code and other coding agents working knowledge of Materialize patterns and troubleshooting playbooks.
mz-deploy. A new Rust-based CLI that lets you define sources, views, indexes, clusters, and other Materialize objects as code. Projects compile locally with no running Materialize instance, so you can run unit tests, inspect query plans, and validate changes in a sandbox before touching a shared environment. Engineering teams and coding agents can now ship Materialize changes the same way they ship application code.
Single Sign-On on Self-Managed. Self-Managed Materialize now supports SSO via OIDC-compliant identity providers like Okta, Microsoft Entra ID, Auth0, and Keycloak, in public preview, so you can manage and provision users through your existing identity provider. Username and password authentication still works alongside it for tools that can't complete an OIDC flow.
We've also shipped several performance and efficiency improvements, including a major CPU reduction on views that use temporal filters (75% down to 4% in our tests), up to 65% faster DDL operations at scale, 17x faster storage usage collection, and ~10% faster materialized view hydration. For the full list of changes, visit the Materialize Release Notes.
📖 Latest from the Materialize Blog
Transaction Processing in the Data Plane
How writing transaction commit logic as a SQL view powered by incremental view maintenance unlocks higher throughput and ~30ms interactive-scale transaction processing in the data plane. Full post >
Building the Live Context Graph for Agents, 28 Weekly Releases Later
How Materialize's 28 weekly releases built the live context graph agents need: fresh data, millisecond queries, and an MCP server so agents can observe, act, and confirm in a tight feedback loop. Full post >
Finding Bugs using LLMs
How Materialize uses LLM-based coding agents to automatically scan every PR, commit, and source file — finding hundreds of real bugs that slipped through existing test suites. Full Post >
📅 Upcoming Events
Real-Time at Fenway - Join us for a Red Sox Game
Boston, MA | July 11th, 2026
Get on the list
In case you missed our last webinar, view the recording for “Real-Time Data Products: Building a Live Digital Twin of Your Operations” Recording
Check out our upcoming events and see where the Materialize team will be next: materialize.com/events
Materialize
slack-1
X
social-icons
Materialize, Inc, 436 Lafayette Street, Floor 6, New York, NY 10003, United States
Unsubscribe
Manage preferences
