Agent Gateways: The Emerging Enterprise AI Perimeter
An AI agent gateway is a centralized access layer that sits between AI agents, large language models, and enterprise tools, enforcing governance, security policies, and cost controls while replacing scattered, ad-hoc integrations with a single, auditable control point across production AI workflows. The strategic shift in enterprise AI is clear: the perimeter is no longer only the network or the CI/CD pipeline, but the agent gateway itself. As organisations move from AI pilots to production-scale agentic AI deployments, autonomous agents are increasingly interacting with AI models, enterprise applications, and business data to automate complex workflows. Without a front door that can track who is calling which model, with what tools, and at what cost, AI quickly becomes an ungoverned shadow infrastructure. The winners in enterprise AI will be those that treat agentic workflow control as seriously as they treat identity, compliance, and DevOps.
Harness: Agents Inside the Existing Pipeline Harness
Harness is betting that the safest way to put agents in production is not to replace pipelines, but to insert agents inside the safety harness enterprises already trust. With more than 1,000 enterprise customers, Harness is extending its delivery platform by launching Autonomous Worker Agents that can replace fixed scripts in delivery pipelines with AI agents that reason through deployment, testing, and security scans under existing governance and audit controls. Each pipeline step that used to be deterministic—like deploying to Kubernetes or running a security scan—can now run as an agent instead, gaining flexibility while keeping the same guardrails. These Worker Agents run on delegates inside customer infrastructure and each agent runs in a sandboxed container with restricted file and network access, its own identity and permissions, and is governed by the same policy engine that gates human deployments.
Perforce Intelligence: MCP-Agnostic Control and Natural Language Testing
Where Harness threads agents into CI/CD, Perforce is building an AI agent gateway around the full AI-driven development lifecycle. Perforce Intelligence introduces an MCP-agnostic Agentic Gateway that provides an orchestration layer for the AI-DLC to control token consumption and ensure compliance across Model Context Protocol servers. In a single install, the gateway can integrate Perforce MCPs, creating a centralized enterprise access layer that lets AI agents and tools coordinate and automate workflows across code, IP, data, infrastructure and testing. This is coupled with AI-assisted testing: non-testers describe what they want to validate in natural language through a single chat interface, and AI executes, giving continuous agent-driven validation across functional, performance, and mobile testing. The control story is not cosmetic. One survey of over 900 CEOs found that “80% of CEOs say their role will be at risk if their company fails to deliver measurable business gains from AI by the end of 2026.”
Nutanix Agent Gateway: Cost, Compliance, and a Unified LLM Frontdoor
Nutanix is taking the idea of the AI agent gateway and turning it into a first-class security and cost perimeter for enterprise AI. Nutanix has announced the general availability of Nutanix Agent Gateway as part of Enterprise AI 2.7, acting as a centralised frontdoor that manages interactions between AI agents, large language models (LLMs), and enterprise tools. Serving as a control layer connecting requestors—AI users and agents—to AI models and MCP servers, it applies access control policies and tool-level filtering across agents. It provides AI developers and platform teams with a single control point to govern agent activity, manage access policies, and monitor token consumption across agentic AI deployments. Token observability is not a minor feature: by centralising token usage data across model providers, organisations can monitor usage, allocate costs, and better control AI spending and identify workloads that can be shifted to self-hosted models. Nutanix Agent Gateway enables organisations to apply consistent governance across agentic AI deployments, whether they rely on public cloud-hosted or self-hosted models.
From Experiments to Production: Why Agent Gateways Will Define Enterprise AI
The common thread across Harness, Perforce, and Nutanix is blunt: enterprises will not scale AI agents without firm control layers. As enterprises move from AI experimentation to production-scale deployment, executives and IT leaders are under pressure to deliver measurable business outcomes while maintaining control over increasingly complex and sometimes costly AI ecosystems. The lack of clear visibility into where AI is being used, how it is making decisions, how to optimise costs, and whether outputs meet security, compliance, and quality requirements creates risks that limit AI’s value in software delivery. Harness shows that agents can deploy, test, and scan under the same governance rigor as traditional DevOps pipelines. Perforce is turning policies into continuous enforcement and extending audit trails into the AI-DLC. Nutanix is making LLM management platforms and agentic workflow control inseparable from cost and compliance. The next phase—autonomous software engineering, where a Jira ticket flows to production through a network of agents—is coming, and those agents will only be allowed near production if there is a gateway firmly in charge.
