Harness, an AI-native software delivery platform, announced an integration with Google Cloud’s Apigee API management platform. The joint capability embeds Harness AI Security directly into Apigee’s API control layer, allowing enterprises to discover, monitor, and defend AI agents, services, tools, and Model Context Protocol (MCP) workflows in real time.
As organizations rapidly deploy autonomous AI agents and large language model (LLM) applications, applying conventional application security protocols creates operational friction and critical blind spots. According to the findings of The Harness State of AI-Native Application Security Report, 74% of developers view security controls as one of the biggest barriers to innovation when it comes to AI. However, 62% of security professionals note that development teams fail to handle risks associated with AI-native applications. The new integration solves the problem through automation of security processes within developer processes.
“Organizations are innovating with agents at breakneck speed,” said Rahul Sood, GM of Application Security at Harness. “As they innovate and adopt new capabilities, they are forcing security teams to keep pace. This integration allows customers building on Apigee to seamlessly detect and protect agents as they are being deployed so developers can innovate confidently.”
Also Read: Cyberhaven Expands Data and AI Security Platform with ChatGPT and Claude
End-to-End Governance Across the AI Agent Lifecycle
By situating security controls within Apigee which manages traffic between end-users, foundational models, back-end APIs, and agentic tools organizations can secure new AI interaction patterns without building separate infrastructure. The platform delivers three core security layers:
Automated Agent Discovery: Continuously inventories AI assets, MCP servers, and connected tools as they deploy across the network. Agent Security Posture Management (Agent SPM) maps complex asset dependencies and evaluates organizational risks automatically.
Real-Time AI Firewall: Monitors live runtime activity to enforce security policies against prompt injection attempts, tool misuse, unauthorized agent-to-agent communication, and data exfiltration.
Comprehensive Activity Observability (AgentTrace): Maintains a complete audit trail across user prompts, AI reasoning steps, tool calls, and outputs to evaluate non-deterministic system behaviors.
“Securing AI agents and emerging technologies requires a paradigm shift in how we think about enterprise protection,” said Ritika Suri, Managing Director, AI & Data ISV Partnerships, Google Cloud. “Through our collaboration with Harness, we are bringing robust, agent-aware security to our joint customers without compromising on the agility to deliver solutions that help organizations innovate with confidence.”