Enterprise AI doesn’t happen in a single ecosystem. Different teams choose different technologies based on their business needs, existing investments and preferred development tools. As a result, organizations often operateagents built with IBM, Amazon Bedrock, Microsoft, Google and custom frameworks across their environments.
The challenge is that most platforms provide visibility and management primarily within their own ecosystems. For enterprises adopting multiple technologies, this platform-centric approach can create fragmented inventories, inconsistent controls and limited visibility into the broader agent estate.
With watsonx Orchestrate, enterprises can manage agents built in different environments through a single experience without rebuilding them or standardizing on one platform.
At TechXchange, we’re introducing the watsonx Orchestrate AI Gateway, a single enforcement layer for discovering third-party agents, models and Model Context Protocol (MCP) servers across platforms such as Amazon Bedrock, Microsoft and Google. It governs those assets through consistent observability, evaluation and optimization. Customers can now interact with those third-party agents directly through watsonx Orchestrate chat and channels.
In addition, the watsonx Orchestrate catalog is evolving into a governed catalog that helps teams discover, share and reuse approved agents and tools rather than repeatedly building the same capabilities. The governed catalog surfaces evaluation data at the moment of discovery and is aligned to OWASP, CoSAI and the IBM AI Risk Atlas. For select agents such as Ziemacs, Cognos Partners and others, builders and admins can see performance, reliability and latency metrics. They can also view risk signals, relevant controls and compliance information such as the EU AI Act.
For select MCP servers such as Box, Visier, New Relic and others, the catalog surfaces MCP protocol conformance, tool schema integrity, tool description safety, input validation and default credential rejection. These capabilitiestransform the catalog into a living quality assurance system for finding, trusting and leveraging the right agentic AI assets.
We’re also strengthening the ability to manage the broader ecosystem of resources agents depend on, including models, tools and MCP servers.
Today’s enterprise agents do not simply generate text. They run custom code, invoke third-party tools through the MCP, coordinate multi-agent graphs and evaluate real-time transactions against corporate backends. When AI agents execute dynamic tool code and integrate with external community MCP servers, shared-kernel infrastructure presents significant enterprise risks.
The watsonx Orchestrate Premium Runtime provides a dedicated, multi-backend hosting, isolation and metering layer. It addresses enterprise risk through hardware-virtualized container isolation, zero lateral movement and hardware-bound resource containment.
For enterprises, this capability means greater visibility into existing AI investments, less duplication and a more consistent approach to managing agents as adoption expands. By providing choice without fragmentation, organizations can continue building with the technologies that work best for them while establishing centralized visibility, management and control across their agent ecosystem.