| title | Okta Open Source Model Context Protocol (MCP) server | |||||
|---|---|---|---|---|---|---|
| meta |
|
The Okta Open Source Model Context Protocol (MCP) server securely connects AI agents and Large Language Models (LLMs) to an Okta org. This abstraction layer enables AI agents and LLMs to interact with Okta admin management APIs using natural language commands.
The Okta Open Source MCP Server translates natural language instructions into structured API calls between LLM clients and an Okta org using Okta's Python SDK v3.4.1. This approach reduces security risks and complexity when connecting autonomous AI agents to IAM systems. The architecture ensures that AI actions remain secure, properly scoped, and fully auditable.
IT admins, developers, and security professionals use the Okta Open Source MCP Server to automate Okta admin tasks through AI-powered interfaces.
Okta offers two ways to deploy an MCP server: the Okta Open Source MCP Server and the Okta Managed MCP Server. The following section introduces each deployment option and helps you choose the one that fits your environment.
To integrate AI assistants securely with your identity infrastructure, the Okta Model Context Protocol (MCP) server can be deployed using two distinct hosting options: Okta Open Source MCP Server and Okta Managed MCP Server.
You can choose the deployment track that matches your technical requirements:
- Okta Open Source MCP Server: You run the server software locally on your own infrastructure. This track is best for developers who want to modify the underlying server code. See Okta Open Source MCP Server documentation.
- Okta Managed MCP Server: Okta hosts and manages the server in the cloud. Your chat client streams instructions over secure HTTPS, which removes local software installations and hosting fees. This setup simplifies onboarding for nontechnical users, such as Okta Identity Governance (OIG) request approvers, who need a ready-made tool without local configuration. See Okta Managed MCP Server documentation.
The following table compares the two hosting options:
| Feature | Okta Open Source MCP Server | Okta Managed MCP Server |
|---|---|---|
| Where it runs | Your infrastructure (a computer, a company container, or a private cloud). | Okta cloud infrastructure. |
| Setup and operation | Install Python packages, clone repositories, configure tooling, and manage infrastructure (OS patches, updates, dependency management). | No installation required. Connect through an HTTPS endpoint. Okta manages infrastructure, updates, scaling, and security patches. |
| Transport protocol | Uses STDIO. | Uses a secure internet connection (HTTPS). |
| User authentication | Device Authorization code flow (interactive users). | OpenID Connect (OIDC) with Proof Key for Code Exchange (PKCE) for interactive users. |
| Service-to-service authentication | JWT private key (API Services for autonomous agents). | JWT private key (API Services for autonomous agents). |
| Cost model | You manage infrastructure, licensing, and operational costs. | Okta managed cloud service with predictable, consumption-based pricing. |
| Best for | Developers testing code in a sandbox or using a command-line interface. | Help Desk teams, IT admins, and automated workflows. |
The Okta Open Source MCP Server addresses security, automation, and integration requirements.
- Security features: The Okta Open Source MCP Server reduces risks by acting as a security checkpoint for AI-initiated actions.
- Access control: The server enforces the principle of least privilege through scope-based tool loading and pre-defined tool definitions.
- Compliance: The system provides audit trails for compliance and security monitoring.
- Operational efficiency: Admins execute identity operations, such as security playbooks and repetitive tasks, with natural language commands.
- Standardized communication: The server uses a standardized protocol to eliminate the need for custom middleware or integration code.
Note: For a list of operations and tools, refer to the okta-mcp-server repo.
The Okta Open Source MCP Server implements AI agent identity security through the following methods:
The Okta Open Source MCP Server ensures that AI agents possess only the minimum permissions required for specific tasks.
- The server manages authentication through OAuth 2.0 scopes and environment variables.
- The server dynamically registers tools based on the granted API scopes of the Okta admin app, ensuring the LLM only sees authorized capabilities.
- The system blocks unauthorized API calls before those calls reach the org.
The Okta Open Source MCP Server maintains a detailed record of all AI agent interactions to support security monitoring.
- The server creates a complete audit trail of all AI agent activities.
- The system integrates with Okta System Logs for centralized monitoring and security investigations.
- The server provides data to support compliance requirements and internal security audits.
The following process describes how an LLM translates a request and how the Okta Open Source MCP Server executes the API call against Okta APIs.
- User command: A user or AI agent (such as GitHub Copilot or Claude Desktop) sends a natural language command through an MCP-compatible client to perform an Okta task.
- Client routing: The MCP client (which hosts the AI agent) securely routes the command to the active Okta Open Source MCP Server instance.
- Server authentication: The Okta Open Source MCP Server authenticates with the org using the pre-configured credentials, such as Device Authorization Grant or Private Key JWT.
- Scope validation: The Okta Open Source MCP Server validates scope twice. First, it filters tools during sign in. Second, it uses a scope guard decorator to confirm active permissions immediately before executing a command and calling the API.
- Response handling: The APIs return a technical response to the Okta Open Source MCP Server.
- Natural language response: The Okta Open Source MCP Server passes the data to the LLM, which formats the information into a natural language response for the user.
The Okta Open Source MCP Server enables AI agents to perform the following actions through natural language.
| Category | Example operations |
|---|---|
| User management | Create, list, retrieve, update, deactivate, delete users, and manage user profile attributes. |
| Group administration | Create, list, retrieve, update, and delete groups. Add and remove users from groups, and view group memberships. |
| App management | List, retrieve, create, update, activate, and deactivate apps. View app assignments. |
| Policy and security management | Manage authentication policies, password rules, MFA configurations, and device assurance policies. |
| System monitoring and logs | Filter System Logs, generate security audit reports, and query-specific sign-in failures. |
| Brands and customization | Manage brands, themes, custom pages, email templates, and custom domains. |
The Okta Open Source MCP Server supports two authentications flows for different use cases:
This browser-based flow is recommended for interactive use and local development. The Okta Open Source MCP Server initiates a flow where the user signs in through a browser. Upon successful authentication, the Okta Open Source MCP Server exchanges the authorization for a secure access token.
This browserless flow is recommended for automation, CI/CD pipelines, and headless environments. The Okta Open Source MCP Server authenticates using a cryptographic key pair, which enables authentication without user intervention.
To connect an LLM to an org, follow the Okta Open Source MCP Server implementation guide for setup instructions.