Descope
Also known as: Descope, Inc., Descope Agentic Identity Hub, Descope Envoy
Descope governs how AI agents sign in and access company tools, APIs and customer applications. The Agentic Identity Hub issues scoped credentials, records authorization events and stores tokens for connections such as Salesforce and Google Calendar. Enterprise administrators can approve or revoke access for individual agents. Developers can use Descope as the authorization server for an MCP server or set up delegated access for visitors' agents on their own websites.
Descope provides identity and access controls for AI agents that work with employees, customers and their software. The Agentic Identity Hub registers an agent as a client so an organization can recognize it and manage its permissions separately from the person it serves. A policy can decide which scopes a client can request on its own and which scopes someone can grant through consent. A protected API or MCP server validates the resulting token before serving the request. Teams can share user records and policies across resources, and connect cloud workloads that already carry an identity from AWS or Google Cloud.
The hub can also manage the credentials an agent's tools need after they connect. A Connections vault stores OAuth tokens and API keys for a user or tenant and can refresh OAuth tokens. An MCP server can retrieve the scoped credential at runtime to make a call to a service such as Salesforce or Google Calendar. A developer can use the same identity project for a customer facing MCP server, an internal application and a plain REST API.
The Enterprise Agent Governance Suite adds a portal for employee access requests. An employee specifies the resources, connections and scopes an agent needs. An administrator can approve, deny or modify that request and can enable automatic approval under an existing policy. An employee can rotate an active agent's secret in the portal. Revoking the agent removes its client credentials. An audit event carries an agent identifier that makes it possible to tie a token issuance or an access denial to a specific authorization record. The activity dashboard brings together events across MCP servers and connections.
For customer applications, Descope Envoy supplies sign in and consent for personal agents using the Personal Agent Protocol. The application can distinguish an agent acting for someone from the person themselves. Its operator can apply scopes, audit access and remove the agent's access while keeping the person's account active. The Descope platform holds a SOC 2 Type 2 report, ISO 27001 and PCI DSS certification and FedRAMP High authorization.
Growth and Enterprise projects can select a supported data region when a project is created, and Growth adds a HIPAA BAA. The Free Forever tier includes 7,500 monthly active users and 2,000 each of agent consents and tokens. Pro starts at $249 per month billed annually, with 5,000 consents and 5,000 tokens and $0.05 for each one beyond that. Growth starts at $799 per month, and the governance suite comes with Enterprise pricing by quote.
Descope fits application developers and security teams that need to add agent authorization to software they already operate and retain control after an agent receives its first token. Thousands of organizations use the wider Descope identity platform, including GoFundMe, Databricks, GoodRx, Navan and You.com, and Gartner named Descope the only Visionary in its 2026 Magic Quadrant for customer identity and access management.
Vendor details
Canonical URL
https://www.descope.com/
Category
Security / SOC agent
Subcategory
Agent identity, delegated authorization and access governance
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
OAuth, OIDC, MCP, Cross App Access, Personal Agent Protocol and a vault for service credentials including Google Calendar and Salesforce.
In practice
A security administrator approves a request for a read scope on a specific MCP resource and later revokes the agent.
An application team creates delegated consent for customers' personal agents without sharing account passwords.
An MCP server retrieves a scoped Salesforce token for a tenant when its tool handles the request.
Sources & related URLs
Related / legacy domains
Research sources
Agentic Index coverage score
8.5 / 14 capabilities · 61%
| Integrations & Tool Calling | Full |
|---|---|
|
Descope authorizes calls to REST APIs and MCP servers and provides a vault for external OAuth and API key connections. A resource can get a scoped token at runtime for a service such as Salesforce or Google Calendar. The vault stores credentials for a tenant or user and can refresh OAuth tokens. The calling tool uses those credentials for its own request to the external service. Sourcedocs.descope.com/agentic-identity-hub/core-components/connectionsread 2026-11-20 |
|
| Workflow Orchestration | Partial |
|
Identity flows can combine agent identification, consent, policy decisions, token issuance and validation. An administrator can change the screens for a customer's agent consent and set what scopes can be presented. These flows control access to the protected resource. The product has no general agent workflow designer for coordinating the business actions that follow authorization. Sourcedocs.descope.com/agentic-identity-hub/agent-loginread 2026-11-20 |
|
| Knowledge Grounding & RAG | Not documented |
|
A protected resource can validate the agent's scope before exposing records to a tool. The application that operates the resource decides where those records come from and what it returns. The Agentic Identity Hub supplies credentials and policies for that path. It has no index for documents or retrieval system that selects relevant passages for a model to use in an answer. Sourcedocs.descope.com/agentic-identity-hubread 2026-11-20 |
|
| Human Oversight & Guardrails | Full |
|
Enterprise administrators can review requests for an agent's resource scopes, grant types and connections and approve, deny or adjust the requested access. Policies can also permit automatic approval. An authorized employee can manage an active agent in the portal and rotate its secret. Revocation removes the client credentials and stops further activity that depends on them. Sourcedocs.descope.com/agentic-identity-hub/agent-governanceread 2026-11-20 |
|
| Security, Identity & Governance | Full |
|
Agent identities are distinct from the people they serve. Descope can issue scoped tokens based on tenant, user, client and resource policy, then let the resource validate the tokens, with role based access control and fine grained authorization for the people administering the project. A resource can fetch third party credentials from the vault so the agent does not need to hold them. The Descope platform holds a SOC 2 Type 2 report, ISO 27001 certification, PCI DSS certification and FedRAMP High authorization, and signs HIPAA business associate agreements on Growth and Enterprise. Sourcedescope.com/security-complianceread 2026-11-20 |
|
| Observability & Auditability | Full |
|
An Agent ID connects audit events to the authorization record for that agent. Teams can inspect token issuance, policy decisions and credential access across connections and MCP resources. The activity dashboard presents the agent events, and policy denials are written to the audit trail even when there is no gateway. Audit history is kept for one week on Free Forever and one month on Pro and Growth, with longer retention as an add on, and Growth and Enterprise can stream events to external audit connectors. Sourcedocs.descope.com/agentic-identity-hub/core-components/agentsread 2026-11-20 |
|
| Memory & State Persistence | Partial |
|
Persistent agent identities link clients to the records used for policies and audit. A Connections vault stores OAuth credentials by user or tenant, tracks their status and can refresh them between tool calls. A later request can retrieve the credential associated with the same account. The stored state does not include a model's conversation history or knowledge it learns while handling tasks. Sourcedocs.descope.com/agentic-identity-hub/core-components/connectionsread 2026-11-20 |
|
| Deployment & Data Residency | Full |
|
The hosted identity service runs each project and can store project configuration and user data in a selected region for Growth or Enterprise customers. The region is set when the project is created. User and tenant data remain there for that project. Applications integrating the service use the regional identity endpoint and can keep the protected API or MCP server in their own deployment. Sourcedocs.descope.com/management/project-settings/multi-regionalread 2026-11-20 |
|
| Prebuilt Agents / Templates / Packs | Partial |
|
Agent Login configures an OAuth server, the consent screens and a Personal Agent Protocol file for a customer website. The team can change the generated flow after setup. Developers can repeat the setup for different sites using the same protocol. The generated components do not contain a ready agent that performs a business task such as finding products or preparing a support case. Sourcedocs.descope.com/agentic-identity-hub/agent-loginread 2026-11-20 |
|
| Triggers & Channel Coverage | Partial |
|
The identity service evaluates a request when an agent authenticates or asks for a token. It supports registered clients, cloud workload identity and credentials issued at runtime. A user or organization can also request agent access through a portal. These events start identity checks. They do not schedule recurring work for the agent or subscribe it to changes in a business application. Sourcedocs.descope.com/agentic-identity-hubread 2026-11-20 |
|
| Model Flexibility & Routing | Not documented |
|
The identity service checks the client and permissions before a protected tool responds. It can authorize callers that organizations build with different frameworks or model providers, using the same policy elements. Model choice belongs to the agent the organization operates. The hub has no model router that selects a generator or changes it according to request cost or latency. Sourcedocs.descope.com/agentic-identity-hubread 2026-11-20 |
|
| APIs / SDKs / MCP Extensibility | Full |
|
Developers can integrate Descope as the authorization server for MCP and REST resources, use its SDKs to obtain tokens and validate tokens at the resource with the project's keys. Credential APIs also support external connections. This allows teams to keep their resource implementation while configuring identity and authorization centrally for clients built in different environments. Sourcedocs.descope.com/agentic-identity-hub/core-components/mcp-serversread 2026-11-20 |
|
| Testing, Debugging & Optimization | Partial |
|
Developers can connect an MCP inspection client to check discovery and login for a protected server. They can set up a project and test its permission paths with a client that requests particular scopes, then inspect the token and audit results. These checks validate the authorization integration. The hub has no evaluation workbench for measuring task output quality over multiple model runs. Sourcedocs.descope.com/agentic-identity-hub/core-components/mcp-serversread 2026-11-20 |
|
| Browser / Computer-use | Not documented |
|
Descope Envoy lets a personal agent use delegated sign in for a customer website. The site can grant limited scopes to that agent and can revoke the identity separately from the person's account. Envoy manages authorization for a browser or client operated elsewhere. It supplies no hosted browser that navigates pages, performs clicks or operates desktop software for the agent. Sourcedocs.descope.com/agentic-identity-hub/agent-loginread 2026-11-20 |
|
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
Pricing
$249 per month billed annually
Base subscription and measured identity usage
Included quota
Free Forever includes 7,500 monthly active users, 2,000 monthly active consents and 2,000 monthly active tokens, with no overages. Pro includes 10,000 users and 5,000 each of consents and tokens; Growth includes 25,000 users and 10,000 each.
What is public
Free Forever includes 2,000 monthly active consents and 2,000 monthly active tokens. Pro starts at $249 per month billed annually. Growth starts at $799 per month billed annually.
Billing mechanics
Pro includes 5,000 monthly active consents and 5,000 monthly active tokens. Additional consents and tokens each cost $0.05 per unit.
Overage / add-ons
Paid plans bill $0.05 for each additional monthly active user, consent or token, and $2 per 1,000 extra machine to machine exchanges. Free Forever requires an upgrade to exceed its limits.
Sales call required
No, self serve available
Free / trial
Free Forever with monthly allowances
Lowest paid plan
Pro
Commercial notes
The Agent Governance Suite is available in Enterprise. Growth adds multi region data residency, a HIPAA BAA and a FedRAMP High add on. Early stage startups can apply for a year of Pro at no charge.
Missing data
Enterprise subscription and governance charges are set in the order.
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…
- Anomali — Anomali combines a security data lake, ThreatStream intelligence and…
Alternatives to Descope
Other products that do the work Descope does, and what each one is.
- ZenitySecurity and governance platform for enterprise AI agents
- Horizon3.aiNodeZero autonomous penetration testing across internal
- Infisical
- OktaIdentity platform extended to AI agents
- Qevlar AI
- Dropzone AI