Salesforce has expanded its MuleSoft Agent Fabric platform with automated discovery capabilities designed to help enterprises identify, govern and manage artificial intelligence (AI) agents deployed across multiple cloud and application environments.
The move comes as organisations rapidly scale the use of AI agents, a trend that industry analysts say will accelerate sharply over the next few years. According to International Data Corporation (IDC), the number of actively deployed AI agents worldwide is expected to cross one billion by 2029, nearly 40 times the level projected for 2025.
While enterprises are experimenting with specialised agents across functions such as inventory management, customer service and logistics, this rapid adoption has led to what technology leaders describe as “agent sprawl” — a fragmented ecosystem of AI agents and tools operating with limited visibility, governance and coordination.
Salesforce said the latest enhancements to MuleSoft Agent Fabric are aimed at addressing this challenge by creating a single control plane for AI agents, tools and related metadata, regardless of where they are built or deployed.
Automated discovery across platforms
At the centre of the update are new “Agent Scanners” that can automatically detect and catalogue AI agents running across platforms such as Salesforce Agentforce, Amazon Bedrock and Google Cloud’s Vertex AI. The scanners are designed to continuously monitor linked environments, identify new or updated agents, and capture technical context such as endpoints and intended functions.
For assets that cannot be automatically detected — including bespoke agents and Model Context Protocol (MCP) servers — MuleSoft Agent Fabric offers simplified registration through URLs, along with access to a curated list of public MCP servers from the Official MCP Registry.
Salesforce said the approach removes the need for manual audits and fragmented documentation, enabling enterprises to maintain an up-to-date view of their AI landscape.
From visibility to governance
Beyond discovery, the platform extracts deep metadata about each agent, including its capabilities, underlying large language models and, where available, the data it can access. This information is normalised and mapped to Google Cloud’s agent-to-agent (A2A) protocol card specifications, allowing consistent interpretation across environments.
All discovered and registered assets are synchronised with the MuleSoft Agent Registry, a central catalogue intended to serve as a real-time system of record for AI agents, MCP servers and related tools.
Salesforce said this always-on cataloguing model is critical for security, compliance and operational teams, which often rely on outdated snapshots of AI deployments.
Enterprise adoption and outlook
“MuleSoft is a massive accelerator for our long-term AI roadmap. With AI moving so fast, MuleSoft Agent Fabric provides the framework we need to scale,” Brad Ringer, Enterprise and Integration Architect at AT&T, said in a statement.
Andrew Comstock, Senior Vice-President and General Manager, MuleSoft at Salesforce, said organisations that succeed in the coming decade will be those that can operate across diverse AI ecosystems without losing control.
“The expanded capabilities of MuleSoft Agent Fabric give enterprises the freedom to innovate across any platform while maintaining unified visibility and control,” he said.
With features such as the MuleSoft Agent Visualizer, enterprises can now filter and analyse their AI footprint across clouds, helping CIOs and IT leaders move from experimental deployments to governed, enterprise-wide AI operations.
As companies transition toward what Salesforce describes as the “Agentic Enterprise,” the latest update signals a shift from isolated AI experimentation to structured, scalable and accountable AI adoption.
