Leading the Future of Agentic Enterprise with MuleSoft
In this presentation, Sonia Wajc and Ian Kelly outline MuleSoft’s vision for the "agentic enterprise," a next‑generation approach to building, governing, and orchestrating AI‑driven agents across an organization. They begin by describing the rapid proliferation of agents—both custom‑built and SaaS‑based—and the need for a unified foundation of discovery, governance, and observability. The core of MuleSoft’s solution is the Agent Fabric, which includes an Agent Registry and Scanners for automatic discovery, an Agent Broker that uses LLM reasoning to compose multi‑agent workflows, and robust governance tools (Agent Governance, Trusted Agent Identity, AI Gateway) that enforce policies, data‑privacy, and cost controls. The MCP Bridge converts existing REST APIs into agent‑compatible services, while the Agent Visualizer provides a real‑time map of the entire agent network. To accelerate developer productivity, MuleSoft offers VIBES (a low‑code coding environment), a Diagnostics Agent, and new AI connectors for NVIDIA NIM and Amazon Bedrock, plus MuleSoft Flow for low‑code orchestration inside Salesforce. Real‑world examples—Barco’s 60 % reduction in development time, an education‑sector predictive student‑performance agent, and public‑sector deployments with FedRAMP High, TX‑RAMP, and defense‑grade clouds—illustrate how the platform works in regulated environments. The session concludes with a roadmap of upcoming features (LLM Gateway, broker patterns, IDP for intelligent document processing) and a call to adopt the Agent Fabric to achieve faster time‑to‑value, secure agent interactions, and holistic visibility across the enterprise.
Chapters
Quick forward-looking statement. Just make your purchasing decisions on generally available products because there's forward-looking statements in this presentation.
Yeah, we are going to talk about some exciting things that are coming up and that are not released yet but are planned for release. So please take a look at that forward-looking statement. Go ahead, Sonia. Yeah.
So my name is Sonia Wajc and I'm joined by my amazing colleague Ian Kelly. And today we're going to be talking about MuleSoft and how MuleSoft helps lead that agentic revolution for your enterprise. So all of the tools we have there and MuleSoft Agent Fabric and developer tools as well. So let's dive in. So today's agenda, we're going to be focusing on that agentic enterprise foundation. Then we're going to move to some of our latest innovations from MuleSoft. And dive a little bit into some of those. And then we're going to look at MuleSoft across industries and the public sector. And then last, just a few next steps for more resources to learn more.
And it is awesome that you've stayed here till 4 o'clock for this session. So thank you all very much.
So the era of agentic transformation is here. And everywhere is creating agents that are coming into your systems, whether that's built in-house or SaaS platforms, etc. Agents are proliferating and we're having to find a way to be able to control them and orchestrate and manage them more effectively.
So it's really that same agentic API explosion that we saw about 10 years ago, 15 years ago that's happening in the agent world. It's just so much faster in the agentic world that it truly is exploding from agents that are in existing tools that are agents built by your organization. And all of that is coming together for this moment. And MuleSoft Agent Fabric really helps us pull that together.
Yeah, and it's an exciting challenge with all of this creativity that all of these agents bring. Bring, but finding a way to orchestrate and manage them is essential. And IT is at the foundation of it. We're looking at the foundation of governance, visibility, and extensibility, where we build that foundation so that you can orchestrate and manage your agents effectively. And we've done this before, as Ian mentioned, with APIs. And we're doing it again now with agents. We don't have to reinvent the wheel, we just have to adapt.
And that really comes with a lot of new capabilities and tying those capabilities to many of the things that we have 20 years of experience with. So if you look at our Exchange catalog and the things that we did for APIs on 8,000+ customers, we learned a lot about how to enable self-service on APIs. How to govern those APIs, how to build reuse for APIs to where different teams aren't blind by what is available to them, and they can actually reuse and reconsume those things, improving and increasing value. And also, it's very standards-based, right? Most of the APIs that we build today are REST-based APIs, and OpenAPI or RAML specifications for those. So we've gone from the composable enterprise with MuleSoft and tools to help the composable enterprise into the agentic transformation. And from the bottom up, we're really driving that agentic enterprise by not just having REST APIs and OpenAPI, but also supporting MCP and industry standards that let you create connection to these tools, to these agents, to these skills across your ecosystem. It helps you reuse and discover agents. So if you are built agents, you have many different agents within your organization. They may be re-interacting with others, but you really want to be able to govern those agents, understand what they do, and MuleSoft is at the forefront of that because of all the capability that we've built for 20 years within Exchange. So, go ahead.
And MuleSoft was built for this. It's a natural evolution of our platform for the agentic transformation. And we're excited to have this transformation with MuleSoft Agent Fabric. That's how we're transforming APIs and agents into an orchestrated network that you're able to discover, orchestrate, govern, and observe in a holistic manner. So diving into our discovery, that's where MuleSoft Agent Registry comes in. It's a centralized repository or catalog where you can see all of your agents wherever they are in one place. And we also have Agent Scanners, which is a way to automatically find agents across existing ecosystems like Amazon Bedrock or Microsoft Copilot. And it's a, it's a place where you're able to ingest those into your MuleSoft registry. So you're able to see those composable pieces from across your team.
And this doesn't just become a registry for agents, for your users to actually interact with the UI. It becomes a place where agents look up capabilities of other agents, other tools that are available. So it became— becomes something systematic that lets your organization move a lot faster and with speed and automates all of this capability.
And with our orchestration, we have MuleSoft Agent Broker. It's that intelligent tooling agent which is able to orchestrate other agents to solve complex problems. So it can find different agents for a different domain, figure out what's useful for it, and use those agents to solve these problems. And it uses an LLM to do this reasoning as well. So it automates that complex workflows.
It also provides you the ability to limit what downstream agents can do and make sure that when you call the broker, you're actually getting the most up-to-date agent or the most appropriate agent for that workload, for that activity, for that tool.
And it works in their own domain. So for example, in an HR domain or in a finance domain, you have an agent broker that's able to utilize other agents or even different tools like MCP servers, etc., and it's fed by an LLM and works together to orchestrate those complex problems that it has to solve.
And that agent broker is ensuring that you're not sending privileged data or data that the agent shouldn't have access to, even if the agent wasn't built with that data restriction in mind. It makes sure that your enterprise data, your organizational data, citizen data doesn't get passed over to agents without really thinking about it, just because you're including that agent.
And moving on to the governance portion, we have MuleSoft Agent Governance as well as Trusted Agent Identity. Governance is at the forefront of every interaction. We want to be able to secure agent-to-agent and agent-to-system interactions effectively. And MuleSoft Agent Governance lets us do that. So we can apply policies to secure those interactions effectively. Trusted Agent Identity helps us provide those permissions throughout the workflows, whether that's the permissions you want to give or take away. You have that control with Trusted Agent Identity. And all of this is also with AI Gateway, a way to govern your agent tools and other models from a single control plane. So you're able to control those costs, the rate limits, and those policies there, and as well as MCP servers and LLMs. So 80-Way and MCP are available now, and LLM Gateway to apply the policies for LLMs is coming soon.
Very soon. And it also gives us the ability to manage those costs, the token usage of every agent, The structure of actually what goes in and out of an agent. There's an example that we have on the floor that might be still going by the time that you leave here today that can show you some of the things that we're making sure don't get passed to an employee onboarding agent. For instance, the badging agent doesn't need my Social Security number and my phone number to get me a nice picture of myself and my name on the bottom, whereas the Workday agent to get me benefits and everything else may need my Social Security number, my location, all those other details. The AI Gateway really makes sure that those details aren't flowing in the wrong directions and that all of those interactions are truly governed. MCP Bridge is a way that we have the ability to take all of your existing APIs, all the structure that we talked about of the things that are existing within Exchange already. We can take those APIs and instead of either using the MuleSoft MCP connector, or writing an MCP server to expose that API, MCP Bridge is open for all Flex Gateway customers right now. It's a free service that lets you just turn any of those APIs into an MCP server, expressing all of the tools that might be available for that API. So it gives you a very simple way to reuse your old API into something that your modern agents can use. That's been launched very recently and it gives us a lot of capabilities and a lot of power.
Right, and REST APIs weren't made for agents. So MCP Bridge gives us that gateway to apply policies to other REST APIs and expose them for agents as well. And MuleSoft Agent Visualizer is our observation portal. Portion. It gives us a way to visualize our entire agent network. So we're able to see our agent broker and other agents it's connected to, or MCP tools. We're able to visualize that all in one spot. So it gives us a holistic view and lets us view any gaps that there might be or any considerations into which agents to use and how our brokers are working as well. So it gives us those real-time insights.
And remember, remember, these are not just MuleSoft agents or AgentForce agents. These are agents that use AWS Bedrock. These are agents from cloud. These are agents from all different locations because they can be ingested within the scanner. So when they show up in the registry, they'll actually be ingested by the scanner. When they're sending data on their usage, how many tokens they're taking in, the types of tools that they're exposing, they'll show up in the visualizer. And then we can query that visualizer and have a visual view for both of our business and our IT teams to get a clear understanding of what we're delivering within our Agentic universe.
And Barco is one example of this. They're a leading company in visualizing products for live events to surgical suites, and they have transformed their agent architecture for the agentic era. So they took their previous API sprawl and their interactions there and transformed them for agents by creating agents and having them work with MCP and A2A well. And they were able to see a 60% decrease in development time by using these tools. So let's dive a little bit deeper into their architecture. We can see that they have a digital workforce, so different agents there from AgentForce, Copilot, et cetera. And these work with the headless agents. So these agents have policies and are communicating through an and then they communicate through MCP to the various APIs and inference connectors and vector connectors to apply different technologies into their workforce. So whether that's RAG with a vector connector or adding LLMs for chat functions, et cetera. So throughout this, they're able to transform their entire architecture into an agentic journey.
And really all of this is about supercharging developer productivity as well. So if we go forward, we can talk a little bit about MuleSoft VIBES and what we're actually exposing to customers within that space. So MuleSoft VIBES is giving us a number of different things. It's giving us an ability to build agents, to build APIs, to build tools, and actually utilize that within a Vibe coding environment. It's letting us deliver faster across that entire development lifecycle and really helping us identify things from that catalog. So when I query Vibes and I say, hey, I want to build a new agent for employee onboarding, it will connect to my registry, find the employee onboarding agent, say, hey, it's here, and I have an MCP tool that I can use, and it will start building out that capability. We've also extended VIBES to put an MCP server on top of our platform. So now we have new agents that will give you things like, give me diagnostics on my platform, help me find my critical bottlenecks, whether that be a database or a file or some other interaction, an MQ address that might be misbehaving. Give me an architecture agent that tells me about how my system is architected. To make sure that I'm actually using best practices as I deliver my APIs and my agents. And also give me a structure to where I can query all of those things and make the agents aware of the platform that they're running on so they can even find new ways to become more efficient. So that's what we're delivering with Vibes. And there's been so many different launches just in the last couple of days, and you're going to see some things in the new coming weeks as well.
Right, and we also have MuleSoft Diagnostics Agent, which helps us look at the different issues and troubleshoot them in our workflows to see what is the problem and the steps we can take to solve them ourselves. And so all of these tools, MuleSoft Vibes, et cetera, help us improve that developer efficiency. So you can take months of developer time into weeks or days.
We also have new connectors that were released with NVIDIA NIM that gives you the full power of the— all the NVIDIA models in all the NVIDIA universe. These are actually partner-delivered connectors. So we've worked with NVIDIA to build these connectors out. So as we build up the capabilities, we give you frontier models, we give you the ability to have local LLMs. We're going to talk about some of our other MCP capabilities and ingestion there. But these NVIDIA services really give us the leading-edge accelerated models to actually utilize within our agentic infrastructure. So this gives you the ability with MuleSoft to expose super powerful models, vector stores, AI models to your agents and make them even more powerful.
And we also recently launched Amazon Bedrock Connector, which allows us to add generative AI and foundation models directly in our workflows with MuleSoft AI connectors.
Within MuleSoft for Flow, we're actually putting the MuleSoft capabilities inside of Salesforce. So we're giving you the ability to write flows, write orchestrations, pull together information in the same way that you would as a MuleSoft developer, but inside of Salesforce as an admin or a user. This gives you the ability to unlock all of those different connectors to connect to agents, even for MuleSoft Flow, and actually pull all that together to build visually in a very low-code, no-code environment to take advantage of all of those capabilities that MuleSoft offers, albeit in the Salesforce ecosystem.
MuleSoft Flow is that low-code solution to really connect your data and systems easily with all of these connectors, as Ian mentioned. And we're helping power the agentic enterprise across every industry, whether that's modernizing legacy systems in the public sector, or empowering institutions to transform with agentic, uh, automation systems, or in healthcare, connecting different data and providers to provide that unified view, or even in education with transforming campuses for that unified student information. All of these sectors are able to utilize MuleSoft Agent Fabric as well as their developer tools to really help transform that architecture and make use of the agents that are out there that help improve the connectivity, the creativity, and the development of their solutions.
We had a customer advisory board yesterday, and one of the customers that was there that was really excited about an agent that they had was an education customer here in Maryland. And that customer is actually building an agent and has built an agent that can look at a university student's participation within a course, to understand whether they'll pass or fail that course. That might have been very useful to a lot of us when we were going through school, right? To know really ahead of time if there was some specific indicators that showed, you know, you're struggling a little bit more than you should in this class, and identifying that early, because there's always other options. And the worst scenario is you've spent a whole semester in a class just to retake it again. I don't have any experience with that. So that might be an agent that you see here at Dreamforce and is utilizing all of these capabilities within MuleSoft. So there are some things going on within global public sector that are really using agents to change how people experience things like education. Right now, we have MuleSoft Government Cloud, which is a FedRAMP moderate environment, and we also have MuleSoft Private Cloud Edition, which operates within an air-gapped environment. Many times in Department of Defense or utilities, or where we have data sovereignty or application sovereignty requirements. We will be launching here our FedRAMP High environment here. What does the date say? I don't know what the date says, so I can say it. I believe it's going to be here in the fall. So the great part about that is it gives us a FedRAMP High environment, which is the highest certified environment for for FedRAMP customers that allows them to run their workloads both in the cloud, on their premises, or within RTF, giving them all the features and capabilities of MuleSoft Agent Fabric within that certified environment, giving them all the features of MCP, A2A, all things we've talked about today in that FedRAMP environment. We will also have a defense announcement for a cloud for the defense partners that we have as well. Within this space, our existing MuleSoft customers will also achieve new capabilities, FIPS 140-3, TLS 1.3, and other frameworks. Our FedRAMP High environment will be in US East that will actually give us very low latency access for all of our federal customers in the East Coast and the Beltway here. That access to latest MuleSoft innovations is going to be amazing because we're opening up to an entire universe of certified individuals that cannot work outside of that space. We're giving them all of the industry-leading technologies, and all of those industry-leading technologies have been deployed on our commercial cloud for months. So the amazing part is everything has been vetted, everything has been moved through. This isn't a initial launch or a beta launch. When you get this in FedRAMP High, you'll get solid capabilities, things that you can rely on for your organization or agency delivered without compromise.
Yeah, and these are coming soon. As you mentioned, fall and winter.
Yeah, some exciting things that we will also have within our FedRAMP High environment is that will inherit TX-RAMP. And some of the other certifications that we have, it will be focused on CGIS workloads, all of those types of workloads that typically need a certified environment. And then as we pull those together, you'll see physically isolated environments for our defense customers as well, all with each one of these features and each one of these capabilities that we've talked about today.
And all of this is coming together, as I mentioned, to really discover, govern, and reuse those agents like we did with the API sprawl. And like Ian mentioned, we're building that composable foundation, but with agents now. And we're orchestrating and gaining that visibility. You have that holistic view. You can see everything that's going on. It's no longer that magic black box. We give you the tools to be able to visualize and make decisions based off of your actual data. And then lastly, we're supercharging that developer productivity with tools like MuleSoft VIBES and the AI connectors. We're really driving that developer productivity and reducing that time to value, helping you make more with less and continue to grow with your agentic transformation. So it unlocks that full potential of your agents across your entire ecosystem.
Yeah, and one thing I didn't mention about MuleSoft VIBES is it's actually open and available to connect through other VIBE coding platforms. So you can connect Cloud Code into the MCP for the platform. You can connect Cloud Code into the developer experience. You can connect Cursor into those. So it's a bring-your-own-IDE environment for any of those. So wherever your teams are comfortable, wherever they're working, they can become effective with MuleSoft immediately. They can become effective building agents, securing agents, discovering, orchestrating, governing, and observing those agents.
Absolutely. And we don't stop there. There's constant innovations to our platforms. To Agent Fabric, and to our developer tools, and more. We're continuously evaluating and upleveling different areas, and making sure that we're keeping up with the different innovations that come up. So we mentioned a few things like Diagnostics Agent, some new connectors like the Amazon Bedrock Connector, MuleSoft for Flow, which came out recently. And we're continuing to grow on that. We're going to be adding broker patterns to our agent broker. We're going to be adding that LLM gateway, as I mentioned. All of this is coming soon and coming fast, and it's available to view more on our website.
And I know we've talked about a lot of innovations in this time, but one of the ones I'm actually most excited for on this sheet that is kind of hidden in there is MuleSoft for Flow IDP, that stands for Intelligent Document Processing. So we've moved Intelligent Document Processing also into the Salesforce experience that connects to AWS Textract and into Einstein for Document Templating and Intelligent Document Templating. So the concept of managing OCR templates, understanding what data might exist in documents, there's no need to learn that anymore. It's all powered by AI. and it's doing document processing for you within the flow experience, just like you have intelligent document processing occurring today on our commercial cloud.
Yeah, we're here to build with you and we, we're here to continue to evolve with you and help you transform with your journey through agentic transformation.
