Identity provider for your AI agents

The Descope Agentic Identity Hub is a dedicated identity provider for AI agents and MCP servers. It gives your AI agents scoped, ephemeral credentials to access sensitive resources.  It integrates natively with OAuth and MCP so you can add auth controls without rebuilding your identity stack or keeping up with these evolving protocols.

Traditional identity falls into two categories: human identities and non-human identities (NHIs). Neither approach fits agents well:

• Human credentials (e.g., SSO, passwords) give agents dangerously broad permissions with very limited ways to revoke dynamically.

• Non-human credentials (e.g., API keys, service accounts) are static. An agent can’t request elevated access when a task demands it.

In short, agents need an identity layer that’s based on ephemeral credentials, granular delegated access that can be raised or lowered in real time based on behavior and context, and strong ongoing compliance with protocols like MCP.

No. Descope layers agentic identity on top of your existing IdP (Okta, Azure Entra ID, etc.). If you’re already using Descope for customer identity, the Agentic Identity Hub extends natively, with no additional configuration needed, allowing your customers’ agent to connect securely to your MCP servers and APIs. 

That covers the identity provider you run. If your enterprise customers want to manage agent access to your product from the identity provider they run, Descope supports that flow as well through Cross-App Access (ID-JAG token validation).

Every action an agent takes is logged in the Descope Agentic Identity Hub and fully exportable to any SIEM provider. Each agentic identity is associated with a user, tenant, and relevant metadata, giving you a clear chain of accountability from agent action back to the delegating user.

Agents receive short-lived, scoped credentials tied to the minimum permissions defined by your access policies. Policies can be set per agent, per tool, per tenant, or based on user roles and JWT claims. 

For sensitive operations, the Agentic Identity Hub supports step-up authentication to add an additional verification gate before the agent can proceed. Credentials expire automatically, so no agent retains standing access to downstream resources.

Yes. The Agentic Identity Hub lets you define granular access policies per agent, per tenant, per MCP server, and per tool-level scope to give security teams centralized visibility and control over what agents are authorized to reach.

The Agentic Identity Hub uses OAuth-based consent and delegation:

• The user authorizes the agent through a consent flow

• The agent gets its own scoped credentials linked back to that user

• The credential vault ensures agents never handle raw tokens directly

This allows downstream services to verify both who the agent is acting for and what it’s permitted to do, without the agent ever holding the user’s primary credentials.

Yes. Descope’s Bring Your Own Auth (BYOA) capability lets you keep your existing identity system as the source of truth. Descope will act as the OAuth authorization server for your MCP servers.

Two paths are available for implementation depending on your setup:

• Identity federation: If your IdP supports OIDC or SAML, you configure it as a provider in Descope’s user consent flow. Users authenticate through your system, and Descope issues MCP-ready OAuth tokens with the right scopes and audience.

• Homegrown auth: If your system doesn’t support OIDC or SAML, you can collect credentials or validate an existing session token against your API using a Generic HTTP Connector in the consent flow, then have Descope issue a JWT.

Yes, with Cross-App Access. Each customer connects their own identity provider, such as Okta or Ping, as a trusted issuer scoped to their tenant. Descope validates the incoming ID-JAG token and issues a standard access token. Your customer admins can complete XAA setup themselves in Descope's self-service setup suite, alongside SSO and SCIM.

Yes. Descope provides the building blocks for multi-tenant MCP gateways that front multiple MCP servers behind a single entry point. The architecture is based on three components:

• MCP server definitions: well-known endpoints, flows, scopes, DCR/CIMD

• Per-tenant connections: isolated credentials, endpoints, and scope mappings for each customer

• Policies: rules governing which agents can access which MCP servers and scopes

These components support two common gateway implementation patterns: mixed gateways that combine internal MCP servers with third-party MCP servers, and per-customer gateways where each customer organization gets its own dedicated MCP server with independent configuration.

Policies are evaluated when the gateway issues tokens, so if an agent isn’t permitted for a given server or scope, the request is denied before any tool execution can happen.

Yes. The Descope Agentic Identity Hub has several customers in production like WisdomAI, Daylight Security, and Cequence Security. Descope was also named a Leader in the 2025 Frost Radar for Non-Human Identity (NHI) Solutions, with the analysts noting how Descope was one of the first platforms to treat AI agents as a first-class identity type.