MuleSoft

Agent Fabric has many capabilities, and they summarize into four main outcomes.

Find every agent. Agent Registry is a cross-platform catalog of every agent, MCP server and API in your enterprise. Scanners connect to each platform’s native APIs — AWS Bedrock AgentCore, Azure AI Foundry, Google Vertex AI, Microsoft Copilot and more — pull agent metadata such as model, tools, ownership, version and deployment status, and normalize it into one searchable registry.

Govern any agent. Agent Fabric’s unified AI Gateway is a single control point for every LLM, MCP and agent call. Universal guardrails and policies cover rate limiting, authentication, PII detection and masking, and data access controls — enforced on any agent regardless of the platform it was built on. An agent kill switch stops a rogue agent at the network, credential and identity layer. Trusted agent identity carries user context through the whole call chain, so an agent calling another agent, an MCP server or an API still acts as a specific user.

Connect AI anywhere. MCP support turns existing APIs into agent-ready tools. AI Connectors reach the systems agents need, including SAP, Workday, mainframe and partner data, and A2A support standardizes agent-to-agent communication. Multi-agent orchestration routes work to the best-fit agent, with as much determinism as the process requires.

Control AI costs. Semantic model routing matches every request to the most cost-effective model that can handle it. Prepaid token budgets cap spend by team, app or agent. A unified FinOps dashboard ranks spend across every provider and model, attributed to specific agents, teams and business units.

Scanners connect to each platform directly and pull the metadata — capabilities, model, tools, ownership, version, deployment status — then normalize it into one searchable registry. They run on a schedule, so new and updated agents keep surfacing rather than waiting to be filed. You can review what is discovered before it is promoted into the governed registry.

Agents can also be registered manually, so anything a scanner does not yet reach — a platform not covered, or an agent built in-house — still sits in the same registry under the same AI agent governance.

Policy is applied at the communication layer, so a third-party agent is governed without being modified or instrumented. That is what makes Agent Fabric vendor agnostic: the same agentic governance envelope applies whether an agent was built in Salesforce, on a hyperscaler, or inside software you bought.

Identity travels with the call too. When one agent calls another agent, an MCP server or a plain API, it still acts as a specific user — using your existing OAuth setup and identity provider.

Three ways, in the order most enterprises need them.

Visibility. One view of spend and budgets across every model, platform and department. A unified FinOps dashboard ranks spend across every AI asset so the biggest leaks show first, and budget management tracks allocation by organization, department or group in real time. You cannot govern what you cannot see.

Efficiency. Semantic routing matches each request to the most cost-effective model able to handle it, rather than sending everything to a premium model. Semantic caching recognizes when the same question is asked in different words and avoids the round trip. Cost optimization trims bloated context, tool descriptions and responses across your agents and MCPs.

Enforcement. Model Wallet locks in prepaid budgets by team, app or agent, so overspend stops before it starts rather than being flagged afterwards. An agent kill switch severs a rogue agent instantly at the network, credential and identity layer.

Yes. Agent Fabric is generally available today, with new and enhanced capabilities rolling out each month.

Point Agent Fabric at the platforms your teams already use and it registers every agent, MCP server and API it finds, so you have a complete inventory in minutes. Add guardrails, connect the systems your agents need, and set budgets from there, at your own pace.

Take the interactive tour, or talk to your account team.

They do different jobs. Agentforce is for building end-to-end agents: build, run, test and optimize interoperable agents within Salesforce, with full lifecycle coverage, trusted context and controlled behavior.

Agent Fabric is for any agent. It is the vendor-agnostic control plane that manages every agent your business runs, including the ones built on a hyperscaler, embedded in third-party software, or written by your own engineers.

MuleSoft, part of Salesforce, helps organizations connect their applications, data, and systems so they can work together seamlessly. Instead of building custom, one-off connections between every system, which are costly to maintain and hard to scale, MuleSoft lets you build reusable APIs that securely link everything from legacy databases to modern SaaS applications.

Its core platform is built around API-led connectivity, an approach that structures APIs in reusable layers so your integrations get faster and easier to maintain over time, not more complex.

Beyond building integrations, MuleSoft provides full API management capabilities to support APIs across their entire lifecycle, from design and security to monitoring and discovery, making it easy for teams to build, protect, and reuse APIs at scale.

As businesses increasingly adopt AI agents, MuleSoft's Agent Fabric extends this same foundation to help enterprises find, connect, govern, and optimize AI agents, giving them secure, managed access to the systems and data they need to take action, while keeping IT and security teams in control.

In short, MuleSoft makes it easier for your systems, data, and now AI agents, to work together securely, at scale, and without reinventing the wheel for every new connection.