Agent Gateways Are Becoming The Control Plane For Enterprise AI

The race to control enterprise AI agents just got real. Palo Alto Networks acquired Portkey, a leading agent gateway startup, while Solo.io contributed its own agent gateway to the Linux Foundation — both moves signaling that agent gateways are rapidly becoming the indispensable control plane for enterprise AI.

In a week that reshaped the infrastructure landscape for generative AI, two major players staked their claims. Palo Alto Networks, the cybersecurity giant, quietly closed its acquisition of Portkey, an early mover in the agent gateway space. Separately, Solo.io, known for its API management and service mesh expertise, donated its agent gateway project to the Linux Foundation, making it fully open source. Together, these events confirm that agent gateways — middleware that manages discovery, routing, security, cost governance, and observability of AI agent interactions — are consolidating into a must-have category for enterprises deploying AI at scale.

Why now? The explosion of AI agents — from customer service bots to autonomous code assistants — has created a governance nightmare. Enterprises need a single pane of glass to monitor which agents are talking to which models, enforce security policies, track token usage and costs, and ensure compliance with emerging regulations. Agent gateways fill that void, acting as the control plane that orchestrates the chaos. Portkey’s platform, for instance, offers failover routing between LLMs, cost tracking, and prompt monitoring — capabilities that Palo Alto will likely weave into its AI security portfolio. Solo.io’s open-source move aims to build a community standard, echoing what Kubernetes did for container orchestration.

Key details: Palo Alto Networks closed the Portkey acquisition for an undisclosed sum — Portkey had raised $12 million from investors including Accel and Sequoia Capital India. Solo.io’s contribution to the Linux Foundation includes the full source code of its agent gateway, documentation, and governance under the foundation’s neutral umbrella. Both are betting that the Model Context Protocol (MCP) — an emerging standard for agent-to-model communication — will be central to how enterprises manage agent fleets. Industry insiders note that the total addressable market for agent gateways could exceed $5 billion by 2028, as every Fortune 500 company moves to deploy AI agents in production.

Analysis: This consolidation mirrors the early days of API gateways, when AWS and Google raced to own the traffic management layer. Agent gateways add new dimensions — prompt injection protection, model-specific rate limiting, and cost allocation per agent. The open-source route from Solo.io suggests that enterprises want vendor neutrality, while Palo Alto’s acquisition points to the security imperative. Venture capital is already flowing: several stealth startups in this space are raising Series A rounds at valuations above $100 million. The message is clear: whoever controls the agent gateway controls the enterprise AI stack.

Outlook: Expect more M&A as cloud hyperscalers (AWS, Microsoft, Google) and security vendors (Zscaler, CrowdStrike) enter the fray. The Linux Foundation will likely form a working group around the agent gateway to standardize MCP integration. Enterprises should start evaluating agent gateways now to avoid being locked into a single vendor. The next 12 months will determine whether this category becomes a commodity layer or a high-margin control plane — but either way, it’s where the battle for enterprise AI supremacy will be fought.