By NHI Mgmt Group Editorial TeamPublished 2025-09-02Domain: Best PracticesSource: WorkOS

TL;DR: App teams are pushing identity controls deeper into developer workflows as August updates add standalone OAuth for MCP, BYOK support in Vault, disposable email blocking, and OIDC attribute mapping, according to WorkOS. The pattern matters because authentication, key governance, and signup abuse are converging into one operational identity surface.


At a glance

What this is: WorkOS’s August updates focus on OAuth for MCP, vault BYOK, OIDC attribute mapping, and account abuse controls.

Why it matters: These changes matter because identity teams now have to govern user authentication, application secrets, and machine-facing integration paths as one connected control plane.

👉 Read WorkOS's August update on MCP OAuth, BYOK, and identity controls


Context

MCP identity risk is no longer limited to agent tooling design. As developers connect AI-facing services to enterprise data and authentication, the core issue becomes how OAuth, token handling, and key custody are governed across application and machine identity flows.

This update set shows the governance gap clearly. Production-ready access patterns now depend on whether teams can authenticate MCP servers, protect vault encryption keys, map external identity claims, and stop low-quality sign-ups without weakening legitimate onboarding.


Key questions

Q: How should security teams govern OAuth access for MCP servers?

A: Treat each MCP server as a delegated identity surface with explicit scope, token lifetime, and revocation ownership. Do not rely on generic app authentication patterns. The important control is whether the server’s access can be reviewed, revoked, and audited as part of the wider machine identity programme.

Q: When does BYOK meaningfully improve vault governance?

A: BYOK helps when the organisation needs independent control over encryption key lifecycle, revocation, and audit evidence. It is most valuable where separation of duties matters and where the enterprise wants a control boundary outside the service provider’s default key management. It does not remove the need for access reviews or onboarding approval.

Q: How do custom OIDC attributes affect identity lifecycle control?

A: Custom OIDC attributes matter because they determine whether provisioning, entitlement assignment, and offboarding receive consistent identity data. If claims are mapped inconsistently, lifecycle automation can misclassify users or break audit trails. The right approach is to standardise the internal profile model before integrating multiple identity providers.

Q: Why do disposable email controls belong in identity governance?

A: Disposable email controls reduce low-quality account creation and help protect free trials and self-service onboarding from abuse. They belong in identity governance because signup hygiene affects downstream access quality, fraud exposure, and support burden. They work best when paired with broader risk signals rather than used as a standalone defence.


Technical breakdown

Standalone OAuth for MCP server access

Model Context Protocol servers need a trusted way to obtain delegated access without embedding credentials in the service itself. OAuth provides that delegation layer by separating user consent, token issuance, and resource access, but the implementation challenge is broader than login. Teams still need token lifecycle handling, scope design, refresh behaviour, and revocation logic that fits the MCP server’s operational model. The real architectural issue is that an MCP server becomes an identity broker as soon as it can reach enterprise data, which means its trust boundary is not just code execution but authenticated downstream access.

Practical implication: treat MCP servers as delegated identity endpoints and govern their OAuth scopes, token lifetime, and revocation paths explicitly.

BYOK in vault key wrapping and revocation

Bring your own key changes who controls the root of trust for encrypted vault data. Instead of the service provider solely managing encryption keys, customer-managed keys in AWS KMS wrap the data keys, which gives the enterprise a separate revocation and audit lever. That matters because key custody, auditability, and access shutdown now depend on KMS policy, key state, and administrative separation as much as on the vault application itself. The security model shifts from platform-only protection to customer-directed cryptographic control, which is a governance as well as a technical decision.

Practical implication: align vault encryption governance with KMS ownership, revocation authority, and access review responsibilities before onboarding.

OIDC attributes and identity claim normalisation

OIDC attribute mapping is the layer that turns external identity provider claims into usable application profile data. When claims are non-standard, custom mapping prevents brittle integrations and reduces the temptation to hard-code assumptions about field names or identity formats. The technical risk is not just failed provisioning, but inconsistent identity records that distort downstream authorisation, lifecycle workflows, and audit trails. Standardising mappings for fields like email, idp_id, first_name, and last_name creates a cleaner identity signal for the application layer, especially when enterprises federate across multiple providers.

Practical implication: validate claim mapping logic against provisioning, access, and audit requirements instead of only testing sign-in success.


NHI Mgmt Group analysis

MCP authentication is becoming an identity governance problem, not just an integration problem. Once an MCP server exchanges OAuth tokens for enterprise access, it inherits the same accountability questions that IAM teams already handle for delegated application access. The difference is that MCP now sits closer to AI-assisted workflows, which makes scope design and token containment more sensitive than conventional app auth. Practitioners should treat MCP onboarding as part of the broader machine identity estate.

BYOK shifts vault security from vendor trust to customer-controlled cryptographic governance. Customer-managed keys do not eliminate platform risk, but they do change who can revoke access and who can prove control over encrypted data. That is a meaningful governance boundary for regulated environments, especially where auditability and separation of duties matter. Practitioners need to view BYOK as a custody decision, not a feature checkbox.

Non-standard identity claims are a lifecycle problem when they break provisioning consistency. Custom OIDC mapping is valuable because identity data has to remain coherent across onboarding, access assignment, and offboarding. If profile attributes differ by provider or application, lifecycle controls fragment and downstream governance becomes unreliable. The practical conclusion is that claim normalisation belongs in the identity design, not at the edge of each app team.

Disposable email blocking is a signup governance control, not a complete fraud strategy. Blocking known disposable domains reduces low-friction abuse in free trials and self-service entry points, but it only addresses one part of account misuse. Fraud patterns often adapt quickly, so identity teams should treat this as an intake control that supports, rather than replaces, broader risk scoring and monitoring. The control is useful when its limits are understood.

From our research:

What this signals

This update reinforces a broader programme reality: identity controls are now distributed across app onboarding, machine access, cryptographic custody, and abuse prevention. Teams that still treat those as separate workstreams will miss the interdependence between delegated access, vault governance, and identity data quality.

Credential custody drift: when claims, keys, and tokens are managed by different teams, accountability weakens faster than most governance models assume. That is where standardisation work, not just tooling, becomes the difference between controllable access and fragmented oversight.

For practitioners, the practical signal is to align application identity design with lifecycle controls already used for service accounts and privileged access. The Top 10 NHI Issues is useful here because it frames the recurring patterns behind exposure, sprawl, and offboarding failure.


For practitioners

  • Govern MCP OAuth as delegated application identity Define the scopes, refresh behaviour, token revocation path, and owner for every MCP server that can touch enterprise data. Reclassify those integrations as part of your machine identity inventory rather than ad hoc developer tooling.
  • Separate vault key custody from application administration Require documented ownership for customer-managed keys, including KMS policy review, revocation authority, and break-glass access. Confirm that the team approving vault onboarding is the same team accountable for key lifecycle controls.
  • Normalize OIDC claims before they reach downstream systems Map external identity provider attributes to a stable internal profile model and test the mapping against provisioning, role assignment, and audit events. Avoid allowing each application to invent its own interpretation of core identity fields.
  • Block disposable domains at the entry point Use disposable email filtering as one control in the signup abuse chain, then pair it with velocity checks, reputation signals, and follow-up verification for higher-risk registrations.

Key takeaways

  • WorkOS’s August updates point to the same underlying governance theme: identity, key custody, and abuse prevention are converging into one operational control surface.
  • The most relevant risk is not any single feature, but whether OAuth, BYOK, OIDC mapping, and signup controls are governed with consistent ownership and lifecycle rules.
  • Practitioners should use this as a prompt to tighten delegated access, cryptographic control, and identity data quality before those gaps compound across the programme.

Standards & Framework Alignment

This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.

OWASP Non-Human Identity Top 10 address the attack and risk surface, while NIST Zero Trust (SP 800-207) and NIST CSF 2.0 set the governance and control requirements practitioners need to meet.

FrameworkControl / ReferenceRelevance
OWASP Non-Human Identity Top 10NHI-03OAuth token handling and vault key custody are core NHI governance concerns.
NIST Zero Trust (SP 800-207)AC-6Least privilege and continuous verification apply to MCP token scope and vault access.
NIST CSF 2.0PR.AC-4Identity and access management governs delegated app access and claim normalisation.

Review delegated access and key custody controls for MCP and vault integrations against NHI-03.


Key terms

  • Model Context Protocol: A protocol that lets AI systems and applications connect to tools and data sources through a standard interface. In identity terms, MCP becomes sensitive as soon as it needs delegated access, because authentication, token handling, and scope design determine what the connected service can do.
  • Bring Your Own Key: A key management model where the customer supplies or controls the encryption key used to protect service data. It changes the trust boundary by moving key custody and revocation authority toward the enterprise, which matters for auditability, separation of duties, and data access shutdown.
  • OIDC Claim Mapping: The process of translating identity provider fields into application profile attributes such as email, name, or tenant-specific metadata. It is critical when organisations federate across multiple providers, because inconsistent mapping can break provisioning, authorisation, and audit consistency.
  • Disposable Email Blocking: A signup control that rejects registrations from temporary or throwaway email domains. It reduces low-quality account creation and some forms of abuse, but it is only one intake control and should be combined with broader risk checks for meaningful protection.

Deepen your knowledge

NHI governance, agentic AI identity, and machine identity security are core topics in our NHI Foundation Level course, the industry's only accredited NHI security programme. If you are responsible for identity security strategy or NHI governance in your organisation, it is worth exploring.

This post draws on content published by WorkOS: August updates covering standalone OAuth for MCP, BYOK in Vault, disposable email blocking, and OIDC attribute mapping. Read the original.

NHIMG Editorial Note
Published by the NHIMG editorial team on 2025-09-02.
NHI Mgmt Group — the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org