- Okta signed a definitive agreement to acquire Permiso Security, with closing expected in the third quarter of its fiscal 2027. Okta said the transaction would not affect guidance issued on May 27. [1] - The companies did not disclose the purchase price. TechCrunch reported that the deal is worth just under $200 million and is structured almost entirely in cash. [2] - Permiso monitors activity after authentication, including AI-agent runs, tool calls, data access and interactions with Model Context Protocol servers. [3] - Okta has not disclosed when Permiso’s technology will become available in an integrated product or how the combined offering will be packaged and priced.
[1] Okta has agreed to acquire Permiso Security, adding technology designed to detect suspicious activity after people, applications and AI agents gain access to cloud systems. The transaction is expected to close in the third quarter of Okta’s fiscal 2027, subject to customary conditions, and Okta said it would not affect the financial guidance issued on May 27.[1]
Neither company disclosed the price. TechCrunch, citing a person familiar with the transaction, reported that Okta will pay just under $200 million in an almost all-cash deal. An Okta spokesperson did not dispute the figure but declined to confirm the terms, leaving the valuation and consideration structure unverified by the parties.[2]
Permiso watches what happens after authentication #
Okta’s core products authenticate identities, manage permissions and govern access to applications. Permiso addresses the activity that follows: observing what those identities do across cloud, software-as-a-service and on-premises environments.
Permiso’s Universal Identity Graph links people, service accounts, machines and AI agents to their credentials and activity. Okta says the platform uses more than 2,500 research-derived signals across more than 70 identity partners to identify excessive privileges, anomalous behavior, policy violations and actions with a potentially large blast radius.[1]
That runtime view is central to Permiso’s AI-agent pitch. An agent may authenticate through an approved identity and then call tools, access databases, connect to a Model Context Protocol server or create sub-agents. Permiso says its platform attributes those actions to an initiating identity, evaluates behavior in real time and can revoke access when activity crosses a risk threshold.[3]
Autodesk is among the first enterprises Permiso has publicly identified as deploying its AI-agent runtime security. In a customer testimonial published by Permiso, Autodesk said it uses the platform to discover agents, maintain a registry and monitor agent runs and tool calls across its products, workforce and cloud infrastructure. That statement confirms a commercial deployment but is not independent product validation.[3]
A runtime layer for Okta’s AI-agent strategy #
Okta for AI Agents became generally available on April 30 as a platform for registering agents, linking them to human owners, managing their lifecycles and controlling access to resources. The acquisition is intended to extend that strategy from identity administration and authorization into runtime detection and response.[4]
The maturity of individual Okta for AI Agents components varies. Okta’s current developer documentation labels the product’s API as a beta release, while July release notes list agent-to-agent connections as generally available and AI-agent access-certification capabilities as early access. Okta’s earlier product materials also subjected unreleased features to delivery caveats.[4]
Permiso therefore offers a complementary capability rather than a replacement for Okta’s existing controls. Okta can manage an agent’s identity and authorized connections; Permiso is designed to determine whether its behavior after authentication departs from policy or an established pattern.
Permiso also brings SandyClaw, a sandbox that executes AI-agent skills and prompts in an isolated environment. Permiso says the system records file writes, network calls and attempts to access environment variables, then applies multiple detection engines to identify potentially malicious behavior before a skill reaches a customer environment.[5]
Okta said it plans to unify identity-threat detection and identity-posture management in a single security offering following the acquisition. It has not disclosed an integration schedule, migration plan, pricing model or whether Permiso will remain available as a standalone product.[1]
Competition is converging on continuous identity controls #
The acquisition puts Okta into closer competition with cybersecurity platforms extending identity controls beyond employees. CrowdStrike announced Continuous Identity for AI Agents in June, using technology from its SGNL acquisition to make authorization decisions based on ownership, device posture and real-time risk.[6]
Palo Alto Networks completed its acquisition of CyberArk on February 11 and introduced the Idira identity platform in May. Idira combines privileged-access, machine-identity and agentic-identity controls, with some capabilities included for existing CyberArk customers and others requiring additional licenses.[7][8]
Permiso emerged from stealth in 2022 with a $10 million seed round and raised another $18.5 million in 2024 in a round led by Altimeter Capital. At the time of the Series A, the startup said it had closed multiple six- and seven-figure licenses with Fortune 500 customers, but it has not publicly disclosed revenue, annual recurring revenue or a customer count.[9]
The reported purchase price would be modest relative to Okta’s balance sheet, though the company has not confirmed it. Okta reported $762 million in cash and cash equivalents and $1.83 billion in short-term investments as of April 30. Revenue for the quarter was $765 million, up 11% from a year earlier.[10]
Companies mentioned #
Further sources #
[[1] Okta, “Okta signs definitive agreement to acquire Permiso Security,” July 30, 2… ↗](https://www.okta.com/newsroom/press-releases/okta-signs-definitive-agreement-to-acquire-permiso-security/)
[[2] TechCrunch, “Okta buys AI security startup Permiso — source says for about $200… ↗](https://techcrunch.com/2026/07/30/okta-buys-ai-security-startup-permiso-source-says-for-about-200m/)
[[3] Permiso Security, “Permiso Brings Identity Runtime Attribution to AI Agents wit… ↗](https://permiso.io/blog/ai-agent-runtime-security)
[[4] Okta official product announcement, developer documentation and release notes f… ↗](https://developer.okta.com/docs/api/secures-ai)
[[5] Permiso Security, “Introducing SandyClaw — The First Dynamic Sandbox for AI Age… ↗](https://permiso.io/blog/introducing-sandyclaw-dynamic-sandbox-ai-agent-skills)
[[6] CrowdStrike, “CrowdStrike Unveils Continuous Identity for AI Agents,” June 15, … ↗](https://www.crowdstrike.com/en-us/press-releases/crowdstrike-unveils-continuous-identity-for-ai-agents/)+4 more
The stories that matter, in one email. Free — unsubscribe anytime.