Watch this Ask the Expert session to explore how AI agents can move beyond dashboards to trigger operational outcomes, including predictions, visualizations and workflows.
Watch the Webinar
You will learn how:
- AI grounded in real-time data and models, not hallucinations, instills confidence.
- MCP makes real-time analytics easily accessible to AI agents.
- Natural-language interfaces enable non-technical users to interact with complex operational systems.
The questions from the Q&A segment held at the end of the webinar are below, and the slides from the webinar are attached.
1. What technology/framework are these MCP apps based on?
MCP apps are an extension of the Model Context Protocol (MCP) specification. While MCP tools typically return text, JSON, or images, MCP apps can return a mini web application, typically rendered in an iframe. This enables rich, interactive experiences such as dashboards, maps, forms, and other visual interfaces directly within the AI agent experience.
2. We are planning AI agents to reduce and standardize model risk management tasks. Are you seeing a lot of work being done in this space?
Yes. By introducing MCP tools to LLMs and AI agents, responses can be grounded in trusted data from deterministic systems rather than relying solely on model-generated content. ESP also supports auditability and traceability, allowing organizations to track when calls were made and what decisions were taken. We are seeing significant interest in applying these capabilities to governance, risk management, and other highly regulated workflows.
3. Can ESP and MCP be used or supported in a SAS Enterprise Guide environment?
ESP typically runs alongside SAS Viya, with ESP Studio providing a web-based development environment. While ESP is not part of Enterprise Guide, it can operate alongside existing SAS Enterprise Guide environments as part of a broader SAS architecture.
4. Is it possible to stop streaming data from being visualized in MCP apps? Could continuous streaming create a memory overflow risk?
Yes. MCP apps are highly flexible and can be designed to start, pause, or stop streaming visualizations as needed. Because developers control the behavior of the application, they can implement mechanisms to manage data flow and resource consumption appropriately.
5. Does ESPCR support exporting ESP MCP tools and running containers that host the MCP server independently?
Yes. ESPCR (ESP Container Runtime) allows you to package an ESP project as a standalone container that can run on-premises, in the cloud, or in edge environments. MCP functionality is included within the container, allowing it to serve MCP requests independently without requiring the full ESP Studio or Event Stream Manager stack. The standalone container can be connected directly to AI agents or LLM-based applications.
6. Can we combine this with SAS Retrieval-Augmented Generation Manager (RAM)? What would that look like?
Absolutely. SAS Retrieval-Augmented Generation Manager (RAM) supports MCP tools, and MCP tools created in ESP work like any other MCP-compatible service. This allows organizations to combine real-time streaming analytics from ESP with document retrieval and knowledge-grounding capabilities provided by RAM. For example, an AI agent could retrieve information from private documents through RAM while simultaneously calling ESP MCP tools for live operational data.
7. Where do customers typically start with MCP and ESP?
Most customers start when they identify a need for AI agents to access real-time data or analytical insights during a reasoning process. Common examples include sensor networks, operational monitoring, asset tracking, financial transactions, and predictive maintenance scenarios.
The starting point is usually an existing ESP project that processes streaming data. Enabling MCP support is straightforward and allows AI agents to access real-time events, analytics, and model outputs through standard MCP interfaces. Because MCP is becoming an industry standard, these ESP capabilities can be exposed to a wide range of AI agent platforms, including SAS RAM and third-party solutions.
8. How do you ensure AI agents are using trusted, real-time data rather than hallucinating?
ESP operates as a standard MCP service, providing secure access to governed, real-time data through well-defined interfaces. The trustworthiness of the solution comes from both the surrounding architecture and the deterministic nature of the data source.
In production environments, organizations can further constrain agent behavior by requiring agents to use only MCP-sourced data and prohibiting modification of that data during reasoning. This creates a highly controlled and auditable workflow that significantly reduces the risk of hallucinations.
9. What does it take to turn an ESP project into MCP tools?
The process is straightforward. Developers define the input endpoint where requests will be received and the output endpoint where responses will be returned. Once those interfaces are identified, the remaining work involves describing the inputs and outputs according to MCP standards, including parameter definitions and metadata.
In practice, this is largely a configuration exercise performed within ESP, making it easy to expose existing streaming analytics and AI capabilities as MCP tools.
10. How do you operationalize insights into actions (for example, tickets or workflows)?
Traditionally, ESP projects generate alerts that trigger downstream actions, such as notifying operators, creating tickets, or initiating automated responses. MCP extends this capability by providing a standard communication layer between ESP, AI agents, and other enterprise systems.
This allows AI agents to combine insights from ESP with information from other sources, reason across that data, and help orchestrate actions such as creating service tickets, initiating workflows, scheduling maintenance, or recommending operational responses. The result is a more interactive and intelligent approach to turning real-time insights into business actions.
Recommended Resources
SAS® Event Stream Processing
Want more tips? Be sure to subscribe to the Ask the Expert board to receive follow-up Q&A, slides, and recordings from other SAS Ask the Expert webinars.
