Harness Integrates Agentic Security with Google Cloud's Apigee

Traditional application security tools often fail to account for the fluid nature of AI agents, which operate with different logic and broader attack surfaces than standard software. The new integration embeds Harness's intelligence directly into the Apigee API control layer. This allows organizations to monitor AI-native assets, including LLM-powered services and Model Context Protocol (MCP) workflows, as soon as they are deployed.

Rahul Sood, general manager of application security at Harness, stated that the move is designed to let developers innovate at machine speed without bypassing security checkpoints. The system provides an automated inventory of AI assets and enforces runtime policies to prevent prompt injections, unauthorized tool usage, and data exfiltration.

Protecting the Agentic Lifecycle

Beyond basic monitoring, the integration introduces AgentTrace, which provides audit trails for prompts, reasoning processes, and tool calls. This observability is critical for managing non-deterministic systems where understanding the 'why' behind an agent's output is essential for enterprise compliance. Ritika Suri, Managing Director of AI & Data ISV Partnerships at Google Cloud, noted that the collaboration brings agent-aware security to joint customers without sacrificing the agility required for modern development. The integration is currently available for all joint customers, marking a shift toward security that scales alongside rapid AI adoption.