APP INTEGRATIONS · FOR AI AGENTS
Descope Agentic Identity Hub
Give agents policy-controlled access to third-party credentials and delegated integrations.
Descope’s Agentic Identity Hub manages agent identities and Connections that vault OAuth tokens or API keys. Agents and MCP servers can retrieve authorized credentials at runtime through SDKs or APIs, with user and tenant scope determining access.
Your integration checklist
- Capabilities
- What you need from your owner
- Create a project and Connection, configure provider credentials and consent, then define agent identity, tenant membership and access policies. Users must connect accounts that require their consent.
- Authentication
- Connection fetches use a project ID plus an access token or Management Key. Access-token requests evaluate policies; Management Keys bypass those checks and belong only in deliberately privileged backend workflows.
- Your first task
- Connect a test account, fetch a scoped credential through the backend and perform one permitted read. Confirm that an unrelated tenant or disallowed action cannot access the same credential.
- Billing & payment
- The free plan has limits for active users, tenants, consents, tokens and machine exchanges. Paid plans add capacity and features with usage-based overages. Upstream service subscriptions and API charges remain separate.
- Agent payment support
- Not verified. Programmatic usage does not establish autonomous checkout support.
- Operational limits
- Autonomous client-credentials agents can fetch tenant-level tokens only for their associated tenants, not arbitrary user-level tokens. Connections and traditional Outbound Apps are separate console objects. Token access does not expand the upstream provider’s permissions.
ORIGINAL AI AGENT STORE INTEGRATION PLAYBOOK
Keep the caller identity attached to each credential request
Editorial recommendation · Reviewed 2026-11-07
When to choose this service
Choose Descope when agents need managed consent, credential refresh and policy enforcement across external accounts. Reuse an existing adequate identity integration when another vault would add complexity.
Authorized provider sessions and managed credential lifecycle are external dependencies. An agent can write an API client but cannot manufacture a user’s consent or valid third-party access.
Suggested workflow
- Identify whether the task acts for a consenting user or a tenant-owned autonomous agent. Create the Connection and assign the appropriate provider scopes.
- Authenticate the caller and pass that identity through the credential fetch. Keep third-party tokens in the backend or gateway when only a tool result is needed by the agent.
- Exercise a permitted read and a denied cross-tenant request in a test environment. Check that revocation or missing consent produces an actionable reconnect flow rather than broader access.
Acceptance checks before real work
- Successful authentication does not by itself authorize every Connection.
- A Management Key bypasses the policy path used by ordinary access tokens.
- Keep user-level and tenant-level credentials distinct throughout the workflow.
If the workflow fails
When consent is missing, return the provider connection flow to the appropriate user. On expired or revoked access, inspect the Connection status and request renewed consent where needed; avoid switching to an administrative key as an automatic retry.