Okta
Also known as: Okta for AI Agents, Okta Identity Governance, Okta Workflows, Okta Platform
Identity platform extended to AI agents: agents registered as first-class identities with owners, an Agent Gateway that authorizes and logs every tool call, step-up human approval for high-risk actions, time-bound access requests and a kill switch; add-on to Workforce Identity plans.
Okta is an identity platform that extended its workforce identity products to AI agents with Okta for AI Agents, generally available since 29 April 2026. The product discovers agents, including shadow AI detected from OAuth consent grants in managed Chrome browsers, and registers them in Universal Directory alongside workforce users, each with a human owner, a business purpose and its own credentials rather than a shared service account.
It manages agents' connections to authorization servers, vaulted secrets, service accounts, applications and MCP servers, and its Agent Gateway enforces identity, policy and audit on every tool call, logging each call without code changes. High-risk actions such as financial transactions, bulk database exports or permission changes require step-up human approval before they execute; access requests are time-bound with admin approval, certification reviews recheck access, and a programmatic kill switch revokes an agent's access.
Cross App Access issues short-lived, scoped tokens for agent-to-app and agent-to-agent connections. It integrates with Salesforce Agentforce, Amazon Bedrock AgentCore, ServiceNow AI Platform, Microsoft Copilot Studio and Google Vertex AI, and exposes APIs on developer.okta.com, including one for MCP server registration. Okta runs in customer-selected cells across the US, EMEA, Japan, Australia, Canada and India and in FedRAMP High government environments, and holds SOC 2 Type II, ISO 27001, ISO 27017 and ISO 27018 among other attestations.
Vendor details
Canonical URL
https://www.okta.com
Category
Security / SOC agent
Subcategory
Agent identity and non human identity governance
Funding status
Public company, Okta Inc, NASDAQ: OKTA. Operates two platforms, Okta for workforce and non human identity and Auth0 for customer identity.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Okta for AI Agents integrates with Salesforce Agentforce, Amazon Bedrock AgentCore, ServiceNow AI Platform, Microsoft Copilot Studio and Google Vertex AI, and governs agents' connections to authorization servers, secrets, service accounts, applications and MCP servers. Cross App Access issues scoped tokens for agent-to-app connections, audit logs stream to the SIEM, and developer.okta.com documents the APIs, including MCP server registration.
In practice
A security team discovers how many agents are actually running across its SaaS estate, including ones no one registered, and gives each a distinct identity instead of a shared service account.
A platform team scopes authorization per task so an agent can only do the specific thing it was deployed for, rather than inheriting the broad rights of the account it was running under.
A regulated organization runs identity in the Okta government cloud under FedRAMP and StateRAMP authorization while extending the same governance to its agent fleet.
Sources & related URLs
Research sources
Agentic Index coverage score
8.5 / 14 capabilities · 61%
| Integrations & Tool Calling | Full |
|---|---|
|
Okta for AI Agents integrates with Salesforce Agentforce, Amazon Bedrock AgentCore, ServiceNow AI Platform, Microsoft Copilot Studio and Google Vertex AI, and manages agents' connections to authorization servers, vaulted secrets, service accounts, applications and MCP servers, with an Agent Gateway on every tool call and Cross App Access for scoped agent-to-app tokens. Sourceokta.com/blog/ai/okta-for-ai-agents-general-availabilityread 2026-09-28 |
|
| Workflow Orchestration | Partial |
|
Discover, register, scope, enforce and contain run as a fixed identity lifecycle pipeline over the customer's agents, with secure multi agent delegation as an identity feature; Okta does not orchestrate the agents' work and runs no agents of its own. Sourceokta.com/products/govern-ai-agent-identityread 2026-09-28 |
|
| Knowledge Grounding & RAG | Partial |
|
Discovered agents are registered in Universal Directory with a human owner, business purpose and credentials: a named inventory of agents, not a retrieval layer grounding them. Sourceokta.com/products/govern-ai-agent-identityread 2026-09-28 |
|
| Human Oversight & Guardrails | Full |
|
"High-risk actions, such as financial transactions, bulk database exports, or permission changes, require step-up human approval before execution", and time bound access requests go through admin approval workflows, so a person approves before a customer agent's action commits. A programmatic kill switch revokes access. Sourceokta.com/products/govern-ai-agent-identityread 2026-09-28 |
|
| Security, Identity & Governance | Full |
|
The trust compliance page lists SOC 2 Type II annually since 2012, SOC 3, ISO 27001, ISO 27017, ISO 27018, CSA STAR Level 2, FedRAMP High and Moderate, DoD IL4 and IRAP. Access to the Okta console runs on standard administrator roles (Super, Org, Group, App, Read-only, Help Desk, API Access Management and others) plus custom admin roles. Sourcetrust.okta.com/complianceread 2026-09-28 |
|
| Observability & Auditability | Full |
|
The Agent Gateway enforces identity, policy and audit for every tool call and logs each tool call without code changes, with a full audit trail and SIEM telemetry streaming: per tool call records of the customer's agents. Sourceokta.com/products/govern-ai-agent-identityread 2026-09-28 |
|
| Memory & State Persistence | Not documented |
|
Identity records and lifecycle state are not agent memory, and the directory is an inventory of agents rather than a store of what they remember; no memory for agents is documented. Sourceokta.com/products/govern-ai-agent-identityread 2026-09-28 |
|
| Deployment & Data Residency | Full |
|
Cells in the US, EMEA, Japan, Australia, Canada and India, with Okta hosting customer data "in data centers selected by Customer", plus FedRAMP High and DoD IL5 government environments supported for Okta for AI Agents: a named region list with customer selection. Sourceokta.com/okta-data-residencyread 2026-09-28 |
|
| Prebuilt Agents, Templates & Packs | Partial |
|
Prebuilt platform integrations (Agentforce, Bedrock AgentCore, ServiceNow, Copilot Studio, Vertex AI) and the Blueprint for the Secure Agentic Enterprise framework are packaged, but no prebuilt agents, policy templates or packs for agent governance are published. Sourceokta.com/blog/ai/okta-for-ai-agents-general-availabilityread 2026-09-28 |
|
| Triggers & Channel Coverage | Full |
|
The Agent Gateway authorizes each tool call in real time as it happens, and shadow AI is detected from OAuth consent grants in managed Chrome browsers, so enforcement and discovery are event driven, with no user launch. Sourceokta.com/products/govern-ai-agent-identityread 2026-09-28 |
|
| Model Flexibility & Routing | Not documented |
|
Okta does not run, route or select models; covering agents built on many platforms is not model flexibility. Sourceokta.com/products/govern-ai-agent-identityread 2026-09-28 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
Documented APIs on developer.okta.com, including the Secures AI API for registering, configuring and managing MCP servers and their tools, token exchange for AI agents and workload principal access, alongside the Management API and an open-source Okta MCP server. Sourcedeveloper.okta.com/docs/api/secures-ai/mcp-serversread 2026-09-28 |
|
| Testing, Debugging & Optimization | Not documented |
|
No testing, red teaming or evaluation of the customer's agents is documented; certification reviews validate access, not agent behavior. Sourceokta.com/products/govern-ai-agent-identityread 2026-09-28 |
|
| Browser & Computer Use | Not documented |
|
No browser or computer use by an agent; monitoring OAuth consent grants in managed Chrome browsers is discovery, not an agent operating an interface. Sourceokta.com/blog/ai/okta-for-ai-agents-general-availabilityread 2026-09-28 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Recent platform changes
Several MCP servers can now sit behind one Okta custom authorization server, generally available in Preview orgs. Cross App Access is built in, so the chain from the end user through every agent hop is kept when agents call those servers, and new APIs manage agent connections.
Bears on: MCP / tool calling / API
View sourcePricing
Workforce Identity Starter at $6 per user per month, billed annually with a $1,500 annual minimum; Okta for AI Agents is an add-on with no published price.
Per user per month (Workforce Identity); agent add-on unit not published
What is public
Workforce Identity suite prices per user per month (Starter $6, Core Essentials $14, Essentials $17), annual billing, the $1,500 annual minimum and a 30-day free trial. Okta for AI Agents is listed as an add-on without a price.
Billing mechanics
Per user per month, billed annually, with a $1,500 annual contract minimum; Professional and Enterprise suites and the Okta for AI Agents add-on are quoted.
Cost watchouts
Inference, not stated by the vendor: the agent governance capability is an add-on on top of a suite, so the entry price does not include it.
Variable cost rationale
Suite pricing scales with users; how agent identities are counted for the add-on is not published.
Additional watchouts
Ask how agent identities are counted for the add-on before modeling cost.
Sales call required
Mixed (some tiers require a call)
Free / trial
30-day free trial, no commitment
Lowest paid plan
Workforce Identity Starter, $6 per user per month billed annually ($1,500 annual minimum).
Commercial notes
Public company (NASDAQ: OKTA).
Key ambiguities
How the Okta for AI Agents add-on is priced and whether agent identities count as users.
Missing data
Okta for AI Agents add-on price and unit; Professional and Enterprise suite prices.
Related vendors
- 7AI — Agentic SOC from the Cybereason founders: domain specialized agents…
- Abnormal AI — Behavioral AI email security with three named autonomous agents: a…
- AirMDR — AI-native MDR whose AI analyst investigates every alert and writes…
- Airrived — Agentic OS that unifies SOC, GRC, IAM, vulnerability management, IT,…
- Andesite — Human-AI SOC platform where customers build their own agents and…
- Anvilogic — Agentic security operations platform that works on top of existing…
Alternatives to Okta
The closest documented capability profiles to Okta among security and SOC agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Idira8.0 / 14A lighter documented profile than Okta
- Horizon3.ai8.5 / 14Adds documented Memory & State Persistence
- Operant AI7.5 / 14A lighter documented profile than Okta
- Qevlar AI8.5 / 14Adds documented Memory & State Persistence
- HiddenLayer9.0 / 14Adds documented Testing, Debugging & Optimization
- Sprinto9.0 / 14Fuller documented coverage on Workflow Orchestration and Prebuilt Agents, Templates & Packs
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded