Hey Data & AI folks,
How the AI Gateway Actually Works (And How to Get Started)
Most enterprises now have agents and assistants that need governed access to Salesforce, SAP, Snowflake, and the dozens of systems behind them. An AI gateway sits between those agents and your data, handling connectivity, identity, and policy in one place. This session explains what a gateway does, where it fits in your architecture, and what changes for your team once it's there.
On November 5, join Foundations for AI, a virtual event where Mike Albritton, SVP of Cloud, and Amit Naik, VP of AI Architecture at CData, break down the gateway piece by piece in our session, Inside the Gateway: Architecture, Components, and Getting Started. They'll cover:
Enterprise MCP: how agents get live access to hundreds of data sources
Governance: what each agent can see and do, and who decides
Cost controls: keeping usage predictable at enterprise scale
Interoperable architecture: the foundation that holds it all together
And, they'll close with a walkthrough of practical tips to get started: connect your systems, set your policies, and put your first agents to work.
Other sessions include:
The AI Gateway Opportunity
The Context Engine
The CIO Perspective: Rolling Out AI at Anaqua
CData Labs: Testing the Gateway in the Open
CData on CData
Visit Foundations for AI to learn more about this free virtual event.
BREAKING from the front row at TiDB, powered by PingCAP SCaiLE 2026!!!! I have been planning to be here for 3 years. Today I am in the first row at the Computer History Museum in Mountain View, and TiDB just made some big announcements. On stage: Max Liu, Co-founder and CEO of TiDB. Vinod Kumar, VP and Head of Engineering at Atlassian. The Head of AI at Plaud. Patrick Barason on the product lineup.
The number that stopped the room:
- TiDB Cloud went from tens of thousands of instances to 3 million+ in 6 months. More than 99% of them were created by AI agents.
Here is everything announced:
- TiDB X goes open source
Core code ships by end of 2026. Compute split from storage cuts costs up to 10x. Compute split from compute cuts costs up to 100x. P99 latency down 65%+. Every new TiDB Cloud cluster now runs on it.
- TiDB Cloud Premium
GA announced, available in November. Isolated compute, 99.99% uptime SLA, 3 zones, customer-managed keys.
- BYOC
Public preview. Your data stays in your own cloud account while PingCAP runs the database.
- TiDB Cloud Filesystem
Public preview. A workspace that outlives the agent’s sandbox. Git resume, checkpoints, forks, and semantic search over files.
- PostgreSQL on TiDB Cloud Starter
Public preview, open to anyone. Vector and full-text search built in.
- TiDB Cloud Essential
Public preview. Set a compute limit and it scales with demand.
The customer proof:
Atlassian consolidated 750+ Postgres clusters into 16 TiDB clusters with 3 million+ tables. They also moved 200+ services off DynamoDB with zero downtime.
Manus AI runs 1.5 million+ TiDB Cloud clusters for apps its agents build.
Kimi runs 1 million+ persistent filesystems and cut data infrastructure costs by an order of magnitude.
Plaud (2.5 million devices, $100 million+ ARR) replaced sharded MySQL and saw 10x throughput.
My take:
I have asked 750+ data and AI leaders what infrastructure looks like when agents do the work. This keynote gave the clearest answer yet. When 99% of your users are agents, speed to provision is the product. Agents try 100 things and keep 1, so cost control decides who wins. And agents forget everything when the sandbox dies. Filesystem is PingCAP’s answer to that.
The open-source TiDB X release is what I am watching next.
The Ravit Show Data and AI, explained through the people building it. 1M+ community | 137K+ newsletter subscribers | 750+ interviews
Please reach out to [email protected] for collaborations.
Best,
Ravit Jain
Founder & Host of The Ravit Show
