Okta teamed up today with some of its customers and partners to showcase the pivotal role ISVs must play in enabling AI agents to work together securely — and introduced an open industry protocol based on existing standards.
At the ‘Identity Summit: Securing Agentic AI’ online event, product leaders, developers, and individuals building AI agents and solutions learned about the challenges enterprises are facing today as they embrace agentic AI. Many organizations are grappling with a lack of control, visibility, and authorization, threatening the promise of productivity gains that can be realized through secure agent-to-agent communication and collaboration.
Okta’s answer to these challenges is the Cross App Access protocol being developed in cooperation with industry-leading ISVs. It’s an extension of OAuth and complements emerging standards like Model Context Protocol (MCP) and Agent2Agent (A2A). Cross App Access shifts control from individual apps to the identity layer, where access policies can be centrally enforced and audited, and enables ISVs to provide seamless cross-app connections for enterprise customers and improve agent access management.
“One of the biggest things we’re hearing from enterprise customers right now is this: they’re excited about what AI can unlock, but they’re also overwhelmed,” says Brendan Ittelson, Chief Ecosystem Officer at Zoom. “Especially when it comes to how AI agents, apps, and systems are all starting to talk to each other. Companies want to scale AI across their businesses, but they’re experiencing friction when it comes to trust and control of how agents and apps connect across systems.”
Okta’s Senior Vice President and Deputy Chief Security Officer, Charlotte Wylie, says what’s happening is profound.
“AI agents are introducing hidden, privileged access that is incredibly difficult to govern, and as CISOs, we’re balancing the immense pressure for AI-driven productivity with the absolute need for security,” she explains.
To truly advance the business, enterprises cannot simply lock down AI, as it will stifle innovation. Instead, the focus should be on working with SaaS vendors to implement open, shared standards that can enable real-time visibility and the effective management of AI access, she says.
“The consequence for us to not do this is quite stark,” she continues. “We will simply not be able to select those vendors that fail to meet the bar. So this matters quite simply because the future of AI depends on trust, and right now, many security teams can’t fully trust agentic AI, which puts the promise at risk.”
Earning that trust will require the use of standards that allow ISVs and SaaS builders to control agent-to-app and app-to-app access in a common way so their platforms and applications can interoperate securely. Current standards were not built to address the unpredictable behavior of autonomous, ephemeral AI agents connecting with other agents. Leveraging existing OAuth extensions, Cross App Access enables an identity provider like Okta to sit in between the OAuth exchanges of two apps or AI-to-app communication. From there, Okta can provide the controls and visibility enterprises require.
Several partners and customers have already expressed support for Cross App Access, including AWS, Boomi, Box, WRITER, and Zoom. It’s Okta’s goal to continue building support for the protocol throughout the industry.
Alyssa Robinson, Chief Information Security Officer at HubSpot, says she wants to be able to demand that ISV partners and SaaS vendors shipping AI-native features provide visibility and granular permissions that put the customer in control of exactly what an AI agent can do.
“We want real transparency,” Robinson says. “We want auditability. We want granular permissions,” adding that without effective control of the actions agents can take, and visibility into the data flows between agents and the data sources they are connected to, “we are going to be in a world of trouble.”
May Habib, CEO and co-founder of WRITER, noted that identity now applies to both people and autonomous agents – and needs to be governed based on goals and outcomes, not just inherited permissions. Agentic AI challenges the very notion of “done,” as behavior emerges and evolves in real-world use. This demands continuous supervision at every layer.
No single company is going to solve these problems alone, notes Habib. Interoperability is table stakes today, but the industry needs to develop the standards that embolden employees to build the agents that transform the way they work.
“We’re very excited about the protocol [Okta is] putting together,” Habib says.
“Where we are enabling these AI agents to act autonomously with no human interaction and… access to multiple systems, this is where we absolutely need our ISV partners and SaaS vendors to be on board with using open standards,” Wylie says, and ensure “that we’re all singing from the same hymn sheet around what protocols we’re using.”
Watch the full on-demand video above, or learn more about how to integrate your enterprise AI tools with Cross App Access on Okta’s developer blog.
