MuleSoft hosted an intimate gathering of practitioners and data/developer advocate executives in New York this week to walk attendees through the evolving API landscape and examine the role of interconnectivity in the new world of artificial intelligence. With the IT industry’s focus skewing towards agent-to-agent (and now agent-to-system) connectivity, MuleSoft went deep on the role of the Agentforce platform (and how MuleSoft for Agentforce is brought to bear) to clarify and validate where its power is capable of surfacing.
What is MuleSoft for Agentforce?
To refresh ourselves, Salesforce Agentforce is the company’s platform for building, designing and managing autonomous AI agents that can support and enhance employee productivity through personalised actions aligned to various business functions. Equally then, MuleSoft for Agentforce works to securely unlock data from any software system or data service to provide Agentforce with “context to reason” so that planning decisions can be made. MuleSoft enables Agentforce to perform actions directly in any system of record or reference to boost productivity and improve customer experiences. We might say that MuleSoft acts as the middleware between data sources and Agentforce AI agents to unlock data and perform actions in other systems.
The company reminds us that MuleSoft Anypoint Platform allows users to develop, manage and secure APIs and integrations for Agentforce from one unified platform.
As Techzine Global has noted before, the capabilities of MuleSoft Agentforce feature technologies that include Topic Center and API Catalog, which allow developers to make their existing APIs accessible to AI agents by adding instructions that help agents understand how to leverage these API connections effectively.
Prompts become code
“Perhaps most intriguing is how MuleSoft now enables the replacement of traditional if-then-else logic with AI agents, essentially turning prompts into code. This approach shines in complex scenarios where simple conditional logic falls short, such as determining what constitutes a ‘good customer’ based on multiple factors and nuanced reasoning,” detailed a report this year, which featured a podcast with Andrew Comstock, senior vice president and general manager of MuleSoft, Salesforce.
Comstock says that as organisations today now work quite directly to incorporate AI agents and assistants into their workflow processes, having a simple approach to integrating their core systems and applications with large language models (LLMs) becomes more critical than ever. He thinks that MuleSoft’s latest product developments enable the company to enable scalable, context-aware AI for all business teams.
Integrating a limitless digital workforce
MuleSoft’s analysis of this market this year has driven the firm to showcase research which suggests that some 93% of enterprise IT leaders have implemented or plan to implement AI agents in the next two years. However, integration challenges hinder companies from fully realising the technology’s potential to create a “limitless digital workforce” today, which can significantly alleviate IT workloads.
The survey analysis results (gleaned from questioning 1,050 enterprise IT leaders) say that 95% of respondents struggle to integrate data across systems. Moreover, only 29% of applications are typically connected within organisations, on average, impacting the accuracy and usefulness of AI agents.
But why does this matter? MuleSoft says it can answer that question.
“[We know that] AI agent outputs depend on connected data that enables a comprehensive understanding of the context and nuances within user queries. These agents gather structured and unstructured data from diverse sources, including CRM, ERP and HCM, as well as email, PDFs, Slack and more, and use it to make decisions and take action for any business process. By integrating their data, systems, and applications, organisations can tap into an AI-powered digital labour force that can act autonomously across their business to successfully carry out both simple and complex tasks,” stated MuleSoft, in its survey presentation.
New this year at MuleSoft Connect AI, featured updates were showcased covering Model Context Protocol (MCP), technology services in MuleSoft Agent2Agent (A2A) connectivity and generative AI tools as well. The recent launch of Salesforce Agentforce 3 includes native MCP support; the company says that this means IT teams can use MuleSoft’s new features to turn APIs into MCP servers and extend Agentforce with new opportunities for agent interoperability.
Agent-to-agent… to agent-to-system
MuleSoft now automates IT issue resolution and enables software engineers to build an observability agent that detects a critical server error in an observability and security platform like Datadog (agent-to-system) and alerts a specialised database triage agent (agent-to-agent).
Engineers can also expand the workflow to enable the triage agent to then query a knowledge base for the error signature, identify the fix and instruct a remediation agent to restart the appropriate cloud service and post a status update to Slack (agent-to-agent to agent-to-system). As businesses adopt more AI agents and the number of applications and AI models they use multiplies, their technology infrastructure becomes increasingly complex and siloed – this is the predicament and reality that MuleSoft now aims to fix.
According to Comstock and team, MCP is rapidly gaining momentum as a standardised, open-source framework for enabling secure bidirectional communication between LLMs or AI agents and diverse external systems. MuleSoft MCP Support enables integrations with LLMs and enterprise systems and provides agents with the ability to take action in any system with governance.
“MuleSoft’s agent orchestration marks the next evolution in enterprise integration. It’s how organisations will power truly intelligent, actionable interactions by unifying APIs, agent actions, governance and multi-agent coordination,” said Comstock.
- MuleSoft MCP Connector – enables IT teams to transform any application or API into an agent-ready asset by bringing MCP to every API.
- MuleSoft for Agentforce: Topic Center – enables IT teams to turn APIs into intelligent actions to empower agents, whether for Agentforce or any other AI agent platform.
- MuleSoft Flex Gateway: MCP Support – enables IT teams to secure and standardise communication with AI agents and models through a unified platform.
New A2A support
The A2A protocol is widely agreed to be growing in popularity as an essential standard for enabling secure, scalable and discoverable communication among multiple AI agents. A2A focuses on direct agent-to-agent collaboration, addressing the need for a common language between diverse AI models.
According to a Google developer blog released in April this year, “A2A is an open protocol that complements Anthropic’s MCP, which provides helpful tools and context to agents. Drawing on Google’s internal expertise in scaling agentic systems, we designed the A2A protocol to address the challenges we identified in deploying large-scale, multi-agent systems. A2A empowers developers to build agents capable of connecting with any other agent built using the protocol and offers users the flexibility to combine agents from various providers.”
MuleSoft A2A Support enables secure A2A communication and discoverability across what the company is calling “the agentic enterprise” today. It creates communication across any agent by exposing them as an A2A server or client. MuleSoft Flex Gateway A2A Support provides a unified gateway to allow agents to find each other to accomplish multi-step tasks quickly.
Generative AI tools
What’s a technology announcement without generative AI these days? Spoiler alert, the answer is obviously nothing. MuleSoft knows this, so the company has detailed new generative AI tools built into Anypoint Code Builder, MuleSoft’s own Integrated Developer Environment (IDE), which lets developers build APIs and integrations faster with natural language prompts and take those capabilities to any AI-first IDE.
Software engineers can accelerate development by automatically aligning work with established format and syntax using natural language to generate API specifications; they can extend MuleSoft development capabilities to any MCP-supported IDE, such as Cursor or Windsurf, using Anypoint Code Builder. They can learn MuleSoft’s proprietary technology with generative Dataweave Transformations via natural language… and, they can also auto-generate standardised API documentation at the time of publication to make it easier for consumers to understand how the API functions.
Evolving industry protocols
If there had been a show t-shirt printed for MuleSoft Connect AI, it might have used the slogan “Integrating With Evolving Industry Protocols”, which might not be catchy, but it does do some way to explaining where MuleSoft is with its approach to orchestrated automated connectivity across standards which appear to be reinventing the IT industry on a monthly basis.
If MuleSoft is capable of following standards across MCP, A2A and all forms of AI in the scalable secure way that it appears to have evidenced so far (and let’s not forget that it has all the scalability yardsticks it needs access to inside the Salesforce mothership), then the scalable orchestration of agent-to-agent and agent-to-system actions with a crucial semantic layer will flourish.