Ask the Expert: How to Supercharge Enterprise Agentic Workflows With SAS Retrieval Agent Manager
Watch this Ask the Expert session to learn how SAS® Retrieval Agent Manager helps enterprises securely deploy and manage AI agents, thereby enhancing decision making and automating domain-specific workflows.
You will learn more about:
- Optimizing knowledge retrieval. Integrate diverse data sources with a human-in-the-loop RAG pipeline for precise, citation-backed responses.
- AI-powered agents. Enable no-code or code-based agents to use plug-and-play generative AI services in creating flexible and scalable agentic workflows.
- Automating secure workflows. Seamlessly connect knowledge, agents and tools to create context-rich automated workflows.
The questions from the Q&A segment held at the end of the webinar are listed below and the slides from the webinar are attached.
Q&A
Is vLLM API based calls also supported by default?
We currently support vLLM via OpenAPI API. We are evaluating a deeper integration.
I assume you support remote MCP servers as well then. Is the support for authentication for the remote MCP server there?
Yes, we will support remote MCP by this quarter. As of now, you could connect to external tools by hosting the MCP server into RAM.
Can you please provide more detail about the security/privacy controls?
In short, from a security and privacy perspective, access to agents and RAG collections in the Retrieval Agent Manager is governed by explicit user permissions. These capabilities are exposed externally only through REST APIs secured with OAuth 2.0. In addition, the platform provides traceability, observability, and auditability by retaining execution and access metadata for a defined period. Using OpenTelemetry, this information can be easily exported and integrated with enterprise monitoring and audit tools.
In the "custom" option, does that mean that it fetches for example the financial docs and stores them locally in the vector DB?
Yes, documents are ingested via custom sources and pushed into the vector database through collections. This process can be automated, so whenever a new document is uploaded, it’s automatically vectorized and added to the appropriate collection, keeping the knowledge base continuously up to date.
Are you exploring integrations with recent developments in MCP like MCP Registry?
Yes, but not directly to the registries cause most of the MCP tool deployed are meant to be used with VSCode or similar for code development or IT action. We could import any containerized MCP tool directly from RAM interface.
What is the difference between NotebookLLM/Coplilot and RAM?
Notebook LLM and Copilot are typical apps used by individual users to summarize documents and generate answers from general knowledge. With RAM, we are trying to build a global solution for your enterprise. When questions come in, there is a specialized agent you build that relies only on data governed by your enterprise and applies analytics or business rules. Then, you produce an answer that you know has been tested, evaluated, and configured in the right way—traceable, consistent, and truly connected to your business data. Instead of just assisting with a response, you get an answer you can actually use to support operational and regulatory decisions.
We have some company internal trained and secured LLM which we can use with API calls. Is it possible to use these kinds of models in SAS RAM?
Yes, our SAS platform has been built to support the OpenAI API specification out-of-the-box, which is typically the standard for deploying these models. However, we are happy to discuss custom solutions for integrating company models. We can manage and use almost any model through API.
What input file formats are supported by SAS RAM?
We support most basic file types, such as PDFs, Word documents, Excel files, and source code. You can see the full table below, which also shows the files for which we support OCR and table extraction. All of this information can be found in the documentation as well (SAS Help Center: Supported File Types)
Is SAS Retrieval Agent Manager available in SAS Viya?
At the moment, SAS Retrieval Agent Manager is a standalone application that can connect and use all the information from Viya. We are working with the R&D department to make this functionality available in Viya. Please stay tuned, as we may have an update soon.
How about CAS datasets?
Yes, this is really the beauty of the MCP and tool calling. We can use it effectively. If you haven't seen it, we can share the answer from GitHub via Salsa. We are deploying an MCP tool to connect with Viya RAM. It is fully capable of connecting with the MCP tool, running, and retrieving data from a CAS table or stored procedures you may have developed. You have full access to everything you've implemented, allowing you to get the answer you need.
Please see additional resources in the attached slide deck.
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.
