tray.ai

Most enterprises are adding AI on top of disconnected systems and wondering why it isn't working. @kchakkarapani, CIO at @Zuora and a Tray customer, wrote the diagnosis in @Forbes: bit.ly/4dkao1T

2012: Nobody knew which APIs were being called or why the bill kept growing. 2026: Nobody knew which MCPs were being called or why the bill kept growing. bit.ly/4wrslma

Something big is coming in June. 👀 We'll be at the Gartner Application Innovation & Solutions Summit to talk all things AI orchestration and MCP. Booth #504. Catch our session, "Avoid losing your job to MCP risk." ⏰ Wednesday, June 03, 2026 bit.ly/4fsQx1s

Tray is hiring. Smart people. Interesting problems. Surprisingly good vibes for an enterprise software company: bit.ly/43aJO4S

This is how MCP sprawl starts. One team ships an MCP server. Then another. Then another. Nobody's tracking what's connected, what's being called, or what it's costing. Sound familiar? Enterprises lived this with APIs and learned the hard way. MCP is next and faster.

Don't let the MCP bomb go off. 💣 In two weeks, learn how to defuse it in a live session on MCP governance. Save your seat: bit.ly/4dDWJlU

You can't govern what you can't see. Introducing MCP Insights in Tray Agent Gateway — usage analytics, cost attribution, identity-stamped execution logs, and audit trails across every MCP tool call. The observability layer enterprise MCP has been missing.bit.ly/4wrslma

"Drops of determinism." That's what it actually takes to make MCP agents reliable. Not better models. Not longer prompts. Business logic that lives in the MCP tool, not the prompt. bit.ly/4d4IWVp

Most enterprises are about to learn that AI agents without orchestration aren't a strategy. They're a mess. Gartner's first-ever Hype Cycle for Agentic AI says it better than we can. Full report: bit.ly/4tvdkgC

The show is called AI: Off Script for a reason. The UFO files dropped last week. We found something significant.... Full show: bit.ly/4cVHTae

MCP adoption is outpacing MCP readiness. Loose standards and minimal governance are fine for experimentation. They're a liability when agents are touching production systems. Learn what enterprise MCP deployments actually need to scale securely: bit.ly/4whqoZv

Nobody knows. That's the problem. When every MCP agent shares a service account, attribution disappears. Something breaks in prod and the answer to "which agent did that?" is a room full of seagulls. Every agent needs its own identity: tray.ai/resources/blog…

Here's where to find us over the next few weeks: Tomorrow — Tracking ROI on AI Agents. LinkedIn Live: linkedin.com/events/7453170… May 28 — Defusing the MCP Ticking Time Bomb. Webinar. June 2 — Gartner App Summit, Las Vegas.

Introducing the SaaSpocalypse Now action figure. Built for the unknown. Agent resilient. Power on demand. Includes emergency fuel. Everything you need to survive. Not sold in stores. But there is a blog: tray.ai/resources/blog…

Live in two days. Episode 2 of AI: Off Script — AI ROI, and Stephen is showing his numbers. May 8 | 2:30 PM ET. 🔗 bit.ly/4tPRLZ7 Until then — Nate on the moment the idea-to-production gap finally closed. Watch the clip.

Something is lurking in your enterprise. It looks like productivity. Your finance team built one last Tuesday. IT found three more this morning. Shadow MCP follows the same arc as shadow IT — but faster and with more at stake: bit.ly/4uub1vb

Most MCP connectors give Claude access to one thing. Some give it access to many. But access isn't the whole story. Tray is now in the Claude Connectors Directory. The logic is defined by you, executed by Claude, exactly as designed every time: bit.ly/4tN47RD

Not everything needs AI. Most automation problems are rule-based or AI-powered. Agentic is powerful — but it's also the most complex and costly to get wrong. New practitioner guide: a framework for classifying any automation problem before you build: bit.ly/42MjIF6

Shadow IT took years to contain. Shadow MCP is already happening. Here's how IT teams are building governance before it gets out of hand.

Speed and velocity aren't the same thing. Most enterprise AI programs are moving fast and going nowhere. This was one of seven things that came out of SaaSpocalypse Now — a panel we hosted with IT leaders from @NotionHQ, @Zuora, and @Yext. Full recap: bit.ly/3Qhowj9

Nobody writes a press release about fixing a missing Salesforce field. But that's where most of the value actually is. Nate walks through a real example. Watch the clip. bit.ly/4mErDOc