Okta expands AI agent access controls with 25 links

Okta has expanded its Cross App Access ecosystem with more than 25 integrations, extending its approach to governing AI agent connections across a wider range of workplace software.

The additions span productivity, design, developer and collaboration tools, including Asana, Atlassian, Canva, Cursor, Figma, Glean, Linear, Slack and Zoom. They are intended to bring agent-to-app and app-to-app connections under Okta's identity controls as companies increase their use of autonomous AI agents in daily work.

Okta is positioning Cross App Access, or XAA, as a way to address a problem many IT teams face when staff begin connecting AI assistants to company systems. Those links often rely on static API keys or user consent flows outside central oversight, leaving administrators with limited visibility into what an agent can access and what actions it takes.

XAA provides a standardised way to route those access requests through identity policies, with actions logged and permissions limited to the scope required for each task. The protocol is built as an extension of OAuth and has been incorporated as an MCP authorisation extension.

More than 25 early adopters are involved in the expansion, spanning software vendors, developer platforms and infrastructure providers. The list includes Anthropic, Aquera, Archestra.AI, Asana, Atlassian, Canva, Cloudflare, Cursor, Datadog, Docker, Figma, Glean, Granola, Keycard, Keycloak, Linear, MintMCP, Scalekit, Serval, Slack, Stytch by Twilio, Supabase, VS Code, WorkOS, Zoom and Zuplo.

Okta groups the integrations into three broad categories: requesting applications, resource applications, and identity infrastructure and gateway providers. The first group includes tools such as Claude, Cursor, Docker, VS Code and Zoom, where users may ask an AI assistant to gather information or complete a task. The second includes systems such as Asana, Atlassian, Canva, Datadog, Figma, Glean, Granola, Linear, Serval, Slack, Supabase and Zoom, which hold the data those agents may need to access.

The third category covers the backend layers that route and govern traffic between agents and applications. Providers in this group include Aquera, Archestra.AI, Cloudflare, Keycard, Keycloak, MintMCP, Scalekit, Stytch by Twilio, WorkOS and Zuplo.

According to Okta, the model is designed to give IT teams a central audit trail of what an AI agent or integrated app does across an organisation, while reducing standing privileges and limiting repetitive consent prompts for end users. It also argues this should reduce the time needed for security reviews when organisations want to deploy agents more widely.

Stephanie Barnett, VP of Presales Engineering, Okta APJ, framed the issue around governance pressures facing companies in the region.

"Right now, organisations across APAC are racing to deploy autonomous AI agents, but they're doing it without the same visibility and control they expect across the rest of their tech stack. Relying on fragmented security models for AI isn't just a headache-it's a massive governance risk. Cross App Access changes the game by giving security teams a unified way to manage identity and access policies, providing total visibility into what agents are doing so organizations can scale AI safely," Barnett said.

Partner backing

Several partners involved in the ecosystem expansion also outlined how they view the role of identity controls as AI tools become more common in enterprise workflows. Their comments focused on reducing fragmented authorisation models and giving administrators a single point of control over agent access.

"As AI workflows become more integrated, interoperability and enterprise security need to advance together. We built Enterprise Managed Auth with Okta to replace fragmented, connector-by-connector authorization with a single, centrally managed model, so admins can extend their existing identity investment instead of provisioning access one connector at a time. The expanding Cross App Access ecosystem means enterprises can scale AI while standardizing on the identity provider already running their environment," said Mayank Malhotra, Product, Anthropic.

Canva linked the approach to how enterprises manage access to its design tools and AI features.

"Canva is already trusted by 95% of the Fortune 500, and our MCP server lets even more teams create, edit and publish on-brand designs with Canva's AI and design tools, all in the same workflow. Enterprise Managed Auth, also known as Cross App Access, makes it clear and simple for enterprises to manage AI access with a system they already trust, enabling teams to create with AI, safely and at scale," said Anwar Haneef, GM and Head of Ecosystem, Canva.

Atlassian emphasised the need for both context and governance when businesses move beyond early AI experiments.

"Enterprises need two things before they can trust AI agents at scale: the organizational context to act effectively and the identity governance to act securely. Atlassian's Teamwork Graph gives agents access to billions of nodes of organizational knowledge, providing the right context to deliver better results with less compute. The XAA protocol available through Okta ensures those connections remain secure and centrally managed, giving enterprises the confidence to move from experimental AI pilots to full-scale business transformation," said Jamil Valliani, Head of AI Product, Atlassian.

Slack focused on access to organisational conversations and reducing repeated sign-ins and authorisation prompts for employees.

"The richest context in any organization lives in its conversations: the decisions, the reasoning, the work in motion. Slack's MCP Server gives AI agents access to that context. And with Cross App Access built into MCP, employees never have to think about it; no authorization prompts, no per-server sign-ins. Organisations govern agent access to Slack once, through the identity policies they already have in place," said Rod García, VP of Engineering, Slack.

Okta also pointed to customer validation through Anthropic's beta programme, where it serves as the featured identity provider for joint customers including HubSpot, Ramp and Webflow. The programme has been used to test XAA-enabled workflows, centralised authorisation and automated offboarding of agent permissions.

Support is also being added at the software development kit level, with official MCP SDKs adopting XAA as the enterprise-managed authorisation extension in TypeScript and Java. Python support is in progress.