The News
TrueFoundry has announced integration with Okta’s Cross App Access (XAA) protocol, extending workforce identity policies to AI agents operating through TrueFoundry’s AI Gateway. The XAA protocol, formally the Identity Assertion Authorization Grant, replaces static API keys with short-lived, task-scoped tokens tied to a user’s active Okta identity. Every agent request flowing through TrueFoundry’s gateway is now evaluated against an organization’s existing Okta policies, creating a unified audit trail across human and machine identities.
Analyst Take
The static API key problem is bigger than it looks
Static API keys are the original sin of agentic security. They don’t expire, they’re rarely scoped to a single task, and when something goes wrong, security teams are left reconstructing access history from incomplete logs. The press release cites industry data suggesting 88% of organizations reported confirmed or suspected AI agent security incidents in the past year, with 97% of breached organizations lacking adequate access controls for their AI systems. Those numbers reflect a structural gap: enterprises built identity infrastructure for humans, then bolted AI agents on top with credentials that belong in the 1990s.
TrueFoundry’s XAA integration attacks this gap at the gateway layer, which is architecturally the right place to solve it. Because every agent request already passes through TrueFoundry’s gateway and MCP orchestration layer, the integration doesn’t require retrofitting individual agent implementations. The identity check happens at the boundary, not inside each application. That’s a meaningful engineering advantage. It means organizations don’t need to rewrite agent logic to enforce policy; they enforce it once, at the control plane.
What ITDMs should actually care about: the governance gap closing
For IT decision-makers, the more important signal here is organizational rather than technical. Most enterprises have spent years building Okta policies for their workforce. Those policies encode role-based access, least-privilege rules, and revocation workflows that passed security review. The TrueFoundry integration means those existing policies now extend to agents automatically, without a separate AI governance framework.
According to ECI Research’s Google GovTech Survey Results, 45.6% of respondents govern AI agent use by permitting it “only for approved use cases under defined policies.” That’s the majority posture: controlled adoption, not prohibition. The challenge is that “defined policies” for AI agents have historically meant custom security reviews built from scratch, precisely because agents didn’t fit into existing identity frameworks. XAA changes that calculus. When agent access runs through the same identity system the security team already trusts, the path from pilot to production gets shorter. That’s not a marketing claim; it’s a procurement and governance reality for any organization trying to scale agentic AI responsibly.
ECI Research’s Google GovTech Survey Results also found that 31.8% of respondents cite “FedRAMP/compliance approval friction for AI vendors” as the single largest blocker preventing widespread AI adoption in developer workflows. While TrueFoundry and Okta aren’t specifically addressing FedRAMP authorization here, the underlying dynamic is identical: AI tools stall at the security gate because they don’t fit existing compliance frameworks. An integration that routes agent traffic through an already-approved identity system could directly reduce that friction, even if it doesn’t eliminate the full compliance burden.
The MCP angle developers should watch
For developers building agentic systems, the XAA integration matters because it’s built on top of OAuth and operates as an authorization extension within MCP. That’s not incidental; MCP is rapidly becoming the de facto protocol for connecting agents to tools and data sources. An identity layer native to MCP means developers can design agent workflows with the assumption that identity and scoping are handled at the infrastructure layer, rather than something each team has to solve independently. Short-lived, task-scoped tokens can also reduce the blast radius when a token is compromised: the damage is bounded to a single task, not an entire system. That’s a fundamental improvement in the security posture of agentic applications.
Looking Ahead
The XAA ecosystem is nascent but directionally correct, and TrueFoundry’s early participation positions it well as enterprise demand for agent governance accelerates. The harder challenge ahead is standardization: XAA needs broad adoption across AI vendors, cloud providers, and MCP server implementations before it becomes the default mechanism rather than an optional integration. Okta’s scale and the protocol’s vendor-neutral design give it a credible path to that outcome, but the timeline depends on how quickly the rest of the ecosystem moves.
For enterprises, the immediate decision is simpler. If you’re running Okta for workforce identity and deploying AI agents in production, the TrueFoundry integration is a low-friction way to close a real security gap without building a parallel governance framework. The fact that it’s included with existing Okta workforce SSO plans removes the budget objection. Organizations that delay will find themselves managing an expanding inventory of static keys and incomplete audit trails, exactly the conditions that produce the incidents the industry is already seeing at scale.
Stay Ahead of Application Development Trends
Get weekly analyst insights, research notes, event coverage, and AppDevANGLE updates delivered directly to your inbox.
Subscribe for Weekly Insights
Join technology leaders, practitioners, and GTM teams following the trends shaping modern software delivery.
Looking for deeper research access?
Explore ECI Research reports, survey insights, and market analysis through the ECI Research Portal.
