Secure AI Connections with Cross App Access and Cloudflare

This is exactly the kind of work that gets me excited about tech alliances. Securing AI agent connections isn't just a nice-to-have — it's table stakes as enterprises scale agentic workflows. Proud to see Cloudflare alongside partners like Okta helping set the standard with Cross App Access. 🔐 Gavin W. C.

AI agents need identity, governance, and security just like human users. Great to see Docker and Okta helping advance secure standards for the next generation of AI-powered workflows. #AI #AgenticAI #Security #Docker #Okta

As agents take on more work, securing how they connect to tools matters more than ever. We're proud to help advance a new industry standard for connecting AI safely as a launch partner in Cross App Access (XAA), alongside 25+ AI leaders including Atlassian, Claude Code, Cursor, Okta, and Zoom. By adopting Cross App Access, our customers get: - AI that plays by their rules - only connecting to what it’s allowed to - Faster launches, without lengthy security reviews that slow teams down  - Smoother user experience, with fewer manual agent connections and fewer authorization prompts Enterprise AI security and productivity are finally aligned.   Read the announcement and see the full list of XAA partner integrations: https://lnkd.in/ggQdHKXe

Big news from Okta and Anthropic: Okta was announced as the first identity provider supporting secure agent connections for Claude. This is an important milestone, but what stands out to me is what it signals for the future of enterprise AI. In nearly every conversation I have with our customers, organizations are looking at how AI agents can drive productivity and automate work across more systems. At the same time, security and IT teams need visibility, governance, and control over what those agents can access and do. That's why identity is becoming one of the most important foundations for AI adoption. Alongside this announcement, our Cross App Access ecosystem continues to grow, with integrations across many of the tools employees are using every day—from Asana and Atlassian to Figma, Glean, Slack, Zoom, and more. As more organizations increase their AI adoption, open standards and secure identity will be critical to enabling innovation without sacrificing security. Excited to see where this goes next! Read more: https://bit.ly/43LCCN9

Excited to share that Okta and Auth0 are advancing the industry standard for secure AI connections – Cross App Access (XAA). With 25+ XAA integrations with some of the industry’s most critical applications, we’re bringing AI tools and their connections to enterprise resources within our identity perimeter. This includes Anthropic's new beta program using Okta to centrally manage authorization for Claude MCP connectors. For B2B SaaS companies and developers, XAA solves a massive hurdle: delivering the strict security controls enterprise buyers demand, without ruining the AI user experience. XAA bridges the gap between AI innovation and enterprise identity, making your applications instantly enterprise-ready. Learn more about how to get started: https://lnkd.in/gFbtwYwt

For enterprise leaders across APJ, the AI conversation has officially shifted. We are no longer debating if we should deploy autonomous AI agents; we are navigating the real-world architectural challenges of managing them. The core issue isn't a lack of innovation. It's that as automated agents connect to apps, pull data, and execute workflows, they easily bypass traditional security guardrails. This is exactly why our latest expansion of the Cross App Access (XAA) ecosystem is a massive milestone for the APJ market. We are leading an open, industry-wide security standard that brings agentic workflows directly inside the enterprise identity perimeter. Through more than 25 new strategic integrations, XAA is a core function of the blueprint for the secure agentic enterprise, to safely govern how AI agents interact with corporate data: Securing the Tools You Use: XAA now natively covers the core platforms where enterprise work happens—including global staples like Slack, Zoom, and Asana, alongside standout Australian success stories like Canva and Atlassian. Agents must evaluate identity and security policies before any data moves. Empowering Developers: Infrastructure and routing gateways like Cloudflare, WorkOS, and Zuplo are adopting XAA. If your teams are building bespoke internal agents, they can plug them directly into Okta’s policy engine from day one. Open Standards, Zero Lock-In: Building on our milestone work with Anthropic and the Model Context Protocol (MCP), this framework remains completely vendor-neutral. You get universal visibility and revocation controls without being tied to a single provider's walled garden. Our research shows that while 91% of organisations in our region are already using AI agents, fewer than 10% have a developed security strategy to manage them. True automation cannot scale in a vacuum. By pairing an independent, neutral identity layer with the world's leading app ecosystem, we are giving businesses the confidence to safely move AI from pilot to production at machine speed. How is your team managing the intersection of AI agents and data access controls? Read more here: https://bit.ly/43LCCN9 #XAA #OktaSecuresAI Dan Mountstephen Arnaud Klineberg Mike Reddie Swan Chin Yeo Bradley Hwang Takashi Watanabe Anand Brahme Matt Porta Ayumi Mifune Marcus, ChenYang Guan Maulik Bawani Heejae Chang

Anthropic Adds Centralized Enterprise Authorization for MCP Connectors 📌 Anthropic has revolutionized enterprise AI connectivity by launching the Enterprise-Managed Authorization extension for MCP. This breakthrough replaces tedious, repetitive user consent prompts with a centralized security framework managed through trusted identity providers. By enabling administrators to enforce global access policies, organizations can now seamlessly and securely scale AI tool integrations across platforms like Okta, Figma, and Atlassian. 🔗 Read more: https://lnkd.in/dDRz8ZPw #Anthropic #Modelcontextprotocol #Identityproviders

Okta for AI Agents went generally available on last week. The product gives autonomous AI agents their own identity, lifecycle, permissions, and audit trail inside enterprise IAM. The architecture says that agents are first-class identities, not bolt-ons. Meanwhile…back in physical security…most physical access control systems (PACs) are still managed through shared accounts, “token” credentials, and manual provisioning workflows… and most have ignored AI Agents all together. The gap between how the enterprise IT world is governing machine identity and how the physical security world is doing it is widening every day. This is true until it’s not and you do have to wonder who and when this inevitable truth will be solved. My guess is by a few in our industry and by many outside of it. More on this in the PhySec Collective slack community. The Access Control Collective The Access Control Executive Brief

Unfortunately, Lee is spot on. Physical security tends to be behind when it comes to what IT considers status quo. Verkada changes that. Jake Leichtling and team are working tirelessly to make access control great (again?). Learn about our latest innovations here: https://lnkd.in/gb5iKYGD

We are opening a small number of Design Partner slots for AgentRail starting in early July. Identity and Access Management was built for humans. AI agents need Runtime Governance. As AI agents gain access to GitHub, Slack, HubSpot, Salesforce, Okta, internal APIs, databases, and enterprise systems, a new challenge emerges: It's no longer just about what an agent knows. It's about what an agent can do. When an AI agent wants to execute a sensitive action, organizations still struggle to answer fundamental questions: → Which agent initiated the action? → On whose behalf is it acting? → What delegation chain led to this decision? → What context was used? → Which policies were evaluated? → Why was the action allowed, blocked, or escalated for approval? → How can the entire decision be reconstructed weeks later? AgentRail is a Runtime Identity & Governance Layer for AI Agents. Before a sensitive action is executed, AgentRail evaluates: • Agent Identity • Delegation Chain • Execution Context • Available Permissions • Applicable Policies • Risk Level Then decides: ✅ Allow ⏸️ Require Human Approval ❌ Block Our goal is to make agentic systems observable, governable, and auditable in production. We're looking for teams already deploying AI agents—or planning to connect agents to business-critical systems. If you're working with technologies such as Dust, Glean, OpenAI Agents, Claude Code, Microsoft Copilot, GitHub Copilot, MCP, agent workflows, or multi-agent architectures, we'd love to talk. We're looking for a small number of design partners to help shape the first version of AgentRail using real-world production use cases. If this resonates with you, let's connect. APPLY HERE : https://lnkd.in/erqNZv7f #AgentRail #AgenticAI #RuntimeGovernance #AIAgents #EnterpriseAI #AIInfrastructure #MCP #MultiAgentSystems #CyberSecurity #AIAct