Palo Alto Networks (NASDAQ: PANW) announced on April 30 its intent to acquire Portkey, an AI Gateway startup that processes trillions of tokens per month for enterprise customers. The deal will integrate Portkey into Prisma AIRS, Palo Alto’s AI security platform, as the centralized control plane for monitoring and securing autonomous agent traffic.

Financial terms were not disclosed. The transaction is expected to close during Palo Alto Networks’ fourth fiscal quarter, ending July 31, according to CRN.

What Portkey Does

Portkey, founded in 2023 by CEO Rohit Agarwal (formerly director of product management at Freshworks), built what amounts to a routing and governance layer for AI agent traffic. The gateway sits between enterprise applications and the LLMs and MCP servers powering autonomous agents, inspecting every transaction, enforcing security policies, and managing failovers.

The platform already serves several Fortune 500 customers and claims 99.99% uptime SLAs for autonomous workloads through semantic routing and automated failovers. It provides access to over 3,000 LLMs and MCP tools through a unified API, with centralized artifact management for versioning and access control across models, agents, and MCP servers.

The Prisma AIRS Integration

Once the deal closes, Portkey becomes the AI Gateway for Prisma AIRS 3.0, which Palo Alto describes as the “industry’s first platform to secure the entire agentic AI lifecycle.” The combined system is designed to enforce three layers of control:

Runtime security. The gateway will inspect AI traffic and enforce governance policies at runtime, scanning agent artifacts and running automated red-teaming against agent behavior. Every agent interaction gets identified, authenticated, and authorized in real time.

Agent identity. Portkey will reinforce Agent Identity Security through CyberArk (which Palo Alto acquired for $25 billion in February), applying least-privilege controls to every autonomous action.

Cost governance. Caching techniques and granular quotas aim to eliminate “bill shock” from uncontrolled agent spending on LLM inference.

“As autonomous agents join the enterprise workforce, they also become a new, unmanaged attack surface,” Lee Klarich, Palo Alto’s Chief Product and Technology Officer, said in the press release. “By integrating Portkey into Prisma AIRS, organizations will be able to confidently deploy and govern AI agents.”

The Acquisition Pattern

Portkey extends a rapid acquisition pace. In the past four months alone, Palo Alto has closed the $25 billion CyberArk deal for identity security, a $3.35 billion Chronosphere acquisition for observability, and a deal for agentic endpoint startup Koi, according to CRN. Each acquisition layers into the same thesis: autonomous agents need purpose-built security infrastructure, and the vendor that controls the gateway controls the market.

The blog post framing the deal cited Gartner research showing 81% of enterprises are either piloting or have fully implemented AI agent solutions, up from experimental levels a year ago.

The Control Plane Race

Palo Alto is not the only vendor betting that AI Gateways become the defining infrastructure layer for agent security. This week alone, Google Cloud formalized its Agentic Enterprise strategy at Cloud Next 2026 with a full governance stack, and CrowdStrike deployed Anthropic’s Claude Opus 4.7 across its security operations platform. The convergence is clear: whoever routes and inspects the agent traffic gets to set the security policies.

Portkey’s integration gives Palo Alto a production-tested routing layer with Fortune 500 scale. The question now is whether enterprises will consolidate agent governance under one vendor’s gateway, or whether the fragmented, multi-vendor pattern that defined cloud security will repeat itself in the agent era.