Tray.ai has unveiled a new capability designed to help enterprises bring order and oversight to the fast-moving world of agent development. The company announced Agent Gateway, a feature within its AI orchestration platform that allows IT teams to create and manage Model Context Protocol (MCP) servers and tools with governance, policies and versioning built in. The move comes as enterprises increasingly grapple with the rise of “shadow” MCP deployments created in programming languages such as JavaScript and Python, often without the visibility or safeguards IT departments require.
The new gateway is aimed at curbing the risks of uncontrolled MCP use while maintaining innovation velocity. It provides a central environment where enterprises can define, test and document MCP services, apply scope and guardrails, and track usage across their agent landscape. According to Tray.ai, this approach mirrors the early years of cloud adoption when APIs proliferated without governance, creating security and compliance challenges that API management platforms eventually resolved. By applying similar oversight to MCP, the company argues, enterprises can mitigate sprawl while supporting the emergence of new interoperability standards such as agent-to-agent (A2A) protocols.
Three key use cases form the backbone of Agent Gateway. Teams can build composite MCP tools using Tray’s Merlin Agent Builder, publish connector-backed MCP tools from Tray’s library of more than 700 integrations, and securely consume external MCP servers with full logging and auditability. Each service is instrumented through Tray Insights Hub to provide an audit trail, which can be streamed to observability tools like Datadog or Splunk. This gives enterprises real-time visibility into agent behavior and ensures that MCP deployments are aligned with company security and compliance policies. Executives at Tray.ai frame the release as an essential step for scaling enterprise use of AI agents.
“Unmanaged MCP use looks a lot like the API sprawl we saw in the early days of the cloud era,” said Alistair Russell, Tray’s co-founder and CTO. CEO Rich Waldron added that the gateway allows teams to reuse trusted tools rather than continually rebuild them, reducing risk and cost while future-proofing operations against evolving standards.
Industry observers note that enterprises are under pressure to innovate quickly while maintaining control. Gartner recently described the challenge as a balancing act between adopting MCP and safeguarding governance. Early adopters, such as a global music retailer that has piloted the system, say the Tray approach allows IT to approve, enforce, version and publish tools securely without slowing delivery.
By extending Tray’s Merlin Agent Builder with governance, control and auditability, the company positions Agent Gateway as a one-stop environment for designing, deploying and maintaining agents at scale. With its multi-protocol design, the feature is built not only for MCP but also for Google’s A2A and future interoperability standards. For enterprises, the promise is a more sustainable model of agent innovation that maintains security, compliance and efficiency as AI adoption accelerates.
