Salesforce Advances Agent Fabric: New Guided Determinism and Governance Controls to Scale Multi-Vendor AI Faster

Expanded agent and MCP discovery, advanced capabilities for deterministic orchestration, and LLM governance give customers a trusted control plane for enterprise-grade agent management

Agentic Enterprises like Capita, Alcon, and Diabsolut are managing and orchestrating hundreds of agents across diverse AI platforms

While the opportunity for AI is clear, learnings from the first wave of adoption have revealed a consistent hurdle: To scale faster, enterprises need more granular control. Not every business process needs to be fully agentic, and not every task should be routed to the same expensive model. Customers want greater choice to optimize these decisions based on real-time performance and cost, but they currently lack the oversight to do so across a fragmented landscape of agents and platforms.

Salesforce today announces a major expansion of Agent Fabric, delivering a trusted agent control plane for your rapidly evolving multi-vendor AI landscape. Agent Fabric now brings automated discovery, intuitive authoring, and centralized LLM governance to your entire Agentic Enterprise. Every handoff, every model selection, every decision - optimized for cost and risk, without sacrificing speed.

Since its launch in September 2025, Agent Fabric has managed and coordinated thousands of agentic instances for customers - from the world's largest enterprises like Capita to specialized industry providers like Alcon and Diabsolut - enabling them to discover, govern, orchestrate, and observe agents across their enterprise with complete interoperability.

"Agent Fabric lets us deploy a coordinated network of agents, quickly and safely. It provides a single governed, agentic layer for our implementation solution. Simply start in Slack to ask for support on projects like workshop planning, user stories, and solution design. The best-fit resources are engaged, whether that's pulling best practices through Certinia and SharePoint MCPs or using Agentforce and other homegrown agents. Now, tasks that previously took days are completed in seconds. Agent Fabric is how we scale AI without losing control." - John Pettifor, SVP, Innovation, Diabsolut

What's New

Shortening the Path to Production for AI:

• Expanded Agent Scanners - Automated discovery now covers MCP servers and new platforms including Amazon Bedrock, Microsoft Foundry, and GoDaddy - accelerating visibility and registration across AI assets with secure OAuth authentication.
• Visual Authoring Canvas - Use a new drag-and-drop interface alongside MuleSoft Vibes to map workflows and human checkpoints, streamlining how developers discover best-fit agents and generate the necessary project scaffolding.
• MCP Bridge - Make existing APIs agent-ready by enabling MCP at scale. And extend enterprise-grade security and rate-limiting without changing code.
• Informatica-hosted MCPs - Pull Informatica's data quality and governance MCP servers directly into your workflows - automatically available in Agent Registry so every agent interaction starts with trusted, governed data.

Bringing Rigor and Oversight to Agent Interactions and Multi-Agent Orchestration:

• Agent Script for Agent Broker - Bring the same guided determinism that powers Agentforce to Agent Broker, letting you define fixed handoff rules while LLMs handle the reasoning in between. Goal-based, autonomous agents paired with trusted, codified workflows result in more consistent and reliable outcomes.

• LLM Governance on AI Gateway - Standardize token management and compliance across your entire multi-LLM stack. Enforce routing rules, unify access, and control costs from a central point, helping to ensure that your data stays safe and your budget stays on track.
• Trusted Agent Identity - Enable agents to execute actions using specific user permissions. For high-stakes tasks like money movement or legal review, receive a mobile approval request, keeping every privileged operation verified and auditable.
• Controlled Registration - Selectively register only the agents and tools that meet your business rules, helping to ensure that teams build with authorized, vetted assets.
• Expanded Model Choice - Use Salesforce's reasoning engine and LLMs, alongside OpenAI and Gemini, to extend Salesforce's built-in trust and data security to interactions across your ecosystem.

Agent Fabric is also available in new regions, including Canada and Japan, and now supports Runtime Fabric deployment, letting you run guardrails directly on your own infrastructure for private cloud and on-premises workloads.

Perspectives

"Agent Fabric is the foundation of our multi-agent evolution - it brings all of our agents under one umbrella, helping them discover each other at runtime and intelligently route tasks based on intent. This is how we will redefine the customer experience and change our business." - Sreenivasa Patibandla, Director, System Integrations & APIs, Alco n

"Operating against a constantly shifting AI landscape is challenging - especially as agent adoption matures. Agent Fabric gives you total oversight over your AI landscape. It allows you to turn a fragmented collection of agents into a unified, high-performance digital workforce." - Andrew Comstock, SVP & GM, MuleSoft

Availability

• Agent Fabric and Infrastructure: Available today in new regions, Canada and Japan, with Flex Gateway support for Runtime Fabric.
• Agent Governance: AI Gateway, MCP Bridge, and Trusted Agent Identity with mobile authorization for high-risk agent actions are generally available today.
• Agent Broker: Beta for deterministic orchestration begins in April 2026. Full GA, including the visual authoring canvas and Salesforce model support, arrives in June 2026.
• Agent Scanners: Support for additional platforms, Amazon Bedrock, Microsoft Foundry, and GoDaddy is available today. Support for MCP servers arrives in May, followed by OAuth in June.

This article may include references to services or features that are still in development and are unreleased. Customers should make their purchase decision based on fully released and available features.

Learn More