Palo Alto Networks builds security layer for Claude

Palo Alto Networks frames the shift less as a new product category than as a change in what security has to watch. A developer using Claude Code can, within its permissions, read private repositories, access secrets, modify files, execute commands and connect to tools through the Model Context Protocol (MCP). Multiply that by thousands of developers, endpoints, skills and integrations, and add employees using Claude.ai and Claude Desktop plus teams building custom agents on the Claude API, and the question moves from which applications an employee can access to what an AI system can see, change and act on. The company's stated approach is to discover, assess, protect and govern AI across its lifecycle, starting with AI coding and extending to knowledge work and autonomous agents. On the discovery side, Cortex Agentic Endpoint Security is meant to surface coding agents, extensions and MCP servers in developer environments, giving security teams visibility into Shadow AI at scale. Before deployment, Prisma AIRS AI Supply Chain Security scans models and agent skills for malicious code and backdoors, while AI Red Teaming tests agent workflows for exploitable weaknesses. At runtime, Prisma AIRS inspects governed prompts through native Claude Enterprise inference hooks, and for Claude Code on developer endpoints it inspects prompts, responses, MCP tool calls and supported network interactions. Agent Identity Security supplies verifiable agent identities and just-in-time, least-privilege access, reducing reliance on standing credentials. Governance sits with Prisma AIRS AI Gateway, which applies centralized policy to LLM, MCP and A2A traffic and enforces quotas and rate limits, with usage analytics showing where AI is consumed and at what cost. The company describes the whole as platformization for the AI era, and its stated reason for the partnership is that frontier AI is changing both how software is defended and how it is built and operated.