F5 integrates AI Guardrails into MuleSoft Agent Fabric

F5 has integrated its AI Guardrails product into MuleSoft's Agent Fabric, and the product is now generally available.

The integration embeds F5's security checks in Agent Fabric's AI workflow layer, allowing organisations to inspect prompts and responses as they move between applications and large language models. It is intended to give security and platform engineering teams a single place to manage policies while keeping existing systems in place.

Businesses deploying AI agents have been grappling with how to monitor and control traffic between users, internal systems and language models. Risks include prompt injection, harmful outputs and the exposure of sensitive information, especially as agents move from pilot projects into customer service, operations and other business processes.

Under the new setup, Agent Fabric's Omni Gateway sends traffic to the F5 AI Guardrails Scan API for inspection before a model is called and before a response is returned. This applies checks inline rather than through a separate security layer, which can add complexity and split monitoring data across multiple tools.

F5 and MuleSoft said the integration is designed to give companies a single control plane for AI traffic policy management. It also supports organisations using Agentforce-based agents, Agent Fabric workflows and custom AI applications that need consistent controls across different systems.

Governance focus

The companies are targeting a governance problem that has emerged as AI tools spread across large organisations. In many cases, security teams have limited visibility into how agentic applications handle prompts, outputs and data, creating potential gaps in oversight.

The controls highlighted include protections against prompt injection, jailbreak attempts, toxic outputs and unauthorised topics. The integration is also designed to reduce the exposure of personally identifiable information and other protected data at runtime.

Another feature is deployment flexibility. Customers can run the product in self-hosted Kubernetes environments, including private virtual private clouds, so prompt and completion data can remain within their own boundaries when data residency or sovereignty rules require it.

The companies also pointed to audit and compliance needs. Decisions made by the system carry telemetry and scan identifiers that can be correlated in the F5 console, which can help security operations teams trace events and support compliance work tied to regulations such as the EU AI Act, GDPR and HIPAA.

Security teams can write and version scanners, blocklists and sensitivity thresholds in the F5 console, with Agent Fabric set to pick up those policy updates dynamically. This is intended to avoid code changes or separate policy rewrites when governance settings are adjusted.

Executive comments

F5 framed the release as a response to growing corporate reliance on AI agents in operational settings rather than experimental use.

"AI agents are moving from experiments into the critical path of the enterprise. The biggest risk in agentic AI is that agents will move faster than enterprise security and governance models can keep up. By integrating the F5 AI Security Platform directly into Agent Fabric, we are putting protection in the path of every prompt and response, where it can operate in real time. That lets organisations move faster with AI while preserving the control, visibility, and accountability their most important business processes demand," said Kunal Anand, Chief Product Officer, F5.

MuleSoft described Agent Fabric as a tool for companies running agents across several models and platforms, and said adding F5 broadens the security options available in that environment.

"We built Agent Fabric as the neutral solution to support enterprises running agents across a mix of models and platforms. By extending Agent Fabric's existing LLM API and AI services to support F5 AI Guardrails as a first-class provider, we're making it even easier for customers to scale their agentic enterprise on the security tooling they know and trust," said Andrew Comstock, Senior Vice President and General Manager, MuleSoft.

The launch reflects a wider shift in enterprise software as vendors add governance and security layers around generative AI tools. As companies push AI agents into more sensitive workflows, suppliers are increasingly trying to position themselves not only around model access and orchestration, but also around the controls needed to monitor data use, apply policy and document decisions.

For F5, the move extends its application and API security business into AI-specific traffic inspection. For MuleSoft, it adds another option for customers using Agent Fabric to connect agents, services and models while keeping policy enforcement closer to where prompts and outputs are exchanged.