Broadcom has released VMware Tanzu Platform agent foundations at the AI in Finance Summit, targeting secure agentic AI operations in enterprise environments.
This new release introduces an agentic runtime that enforces “secure-by-default” principles. It is designed to help organisations transition from isolated AI experiments to production-ready autonomous agent deployments that comply with strict governance and security controls.
Broadcom Tanzu division general manager Purnima Padmanabhan said: “Agentic application development is evolving fast. We are partnering with our customers to innovate through this changing landscape. Tanzu Platform agent foundations give you a quick start to move your agentic ideas into production today on a modern private cloud with VMware Cloud Foundation 9.”
The Tanzu Platform agent foundations run on VMware Cloud Foundation (VCF) and extend established developer workflows to AI use cases. The system provides a platform-as-a-service environment, allowing platform teams to manage AI agents with the same tooling used for business-critical workloads.
The agentic runtime implements a “deny-by-default” model to restrict agent access, governing network ingress and egress as well as resource allocations. It requires explicit authorisation for any interactions with data, tools, or external models.
A key technical component is the use of Buildpacks, a departure from Dockerfiles, to build agent containers. These are automatically patched and verified, reducing the risk profile by ensuring only authorised software enters production, eliminating the need for manual security reviews of container builds.
Agent-to-agent and agent-to-infrastructure credential leaks are addressed through structural secrets isolation, with VMware vDefend providing continuous credential safeguard across both infrastructure and SaaS integrations.
According to Broadcom, zero-trust networking is enforced, preventing agents from initiating unsolicited connections or accessing unauthorised internal resources. All service bindings require explicit approval and are strictly monitored.
Developers can start by using a curated, pre-built agent setup, with IT controlling model access, Model Context Protocol (MCP) server usage, and integration with marketplace services. The environment supports enterprise data engines such as VMware Tanzu for Postgres (with pgvector), alongside data streaming and caching systems, used to facilitate memory and context management for agent workflows.
Operational scaling is achieved via VMware Cloud Foundation APIs, abstracting infrastructure mechanics away from developers and deploying supporting services using VMware vSphere Kubernetes Service.
The platform dynamically scales compute, storage and networking resources in response to agent demands and uses automated failure recovery mechanisms to meet expected uptime targets.
Integrated AI gateways regulate model and tool access, supporting alignment with corporate usage and cost policies, and ensuring compliance with internal safety filters.
Broadcom highlights that deploying agentic AI in enterprise settings depends on having a robust, isolated execution environment that keeps the runtime and agent code separate to maintain stronger security.
The zero-trust approach means agents cannot operate outside their defined service boundaries without explicit permissions, a requirement for regulated sectors such as financial services and government. These restrictions address risks associated with uncontrolled agentic behaviour or lateral movement across environments.
Secrets management uses a credential manager to inject secrets into agent environments at runtime, reducing risks tied to static or hardcoded credentials. This approach limits the surface area for credential compromise during inter-agent or agent-to-service exchanges.
Tanzu Platform agent foundations also incorporate the new MCP Gateway, ensuring all agent actions are routed through a single point of control. This centralisation provides visibility for auditing, incident response, and operational feedback, while enabling connection to both remote and on-premises MCP servers as required by enterprise deployment preferences.
Last month, Broadcom introduced Symantec CBX (Carbon Black XDR), a cloud-based security platform integrating technology from its Symantec and Carbon Black enterprise cybersecurity offerings. The platform aims to meet the needs of organisations lacking fully staffed Security Operations Centres but facing advanced cyber threats.
