Your IdP is the source of truth — PortEden's policies follow it in seconds.
PortEden pulls users, groups, and attributes from Microsoft Entra ID, Okta, Google Workspace, and JumpCloud over SCIM 2.0 and OIDC, then binds them to AI access policy. When your IdP says someone is gone or their role changed, every AI request from Claude, ChatGPT, Copilot, Gemini, or any MCP server reflects that change in seconds — not at the next audit cycle.
Free tier · No credit card · SCIM 2.0 + OIDC + SAML 2.0
Your IdP is current. Your AI access policy isn't.
Identity changes in Entra ID, Okta, or Google Workspace dozens of times a day — joins, departures, role changes, group edits. If the AI access layer doesn't hear about every one of them in real time, the gap between "what your IdP says" and "what AI clients can do" becomes the largest unmanaged risk in your stack.
Stale identity = stale policy = unauthorized AI access
When someone leaves and your IdP deprovisions them, the AI client they were using yesterday still has their token cached. The model keeps answering questions on data they no longer have any right to.
Two systems for identity, two answers
Your IdP says someone moved from Sales to Legal. PortEden still has them on Sales policies for hours or days. The two systems disagree and the audit log can't tell you which one was authoritative when the request fired.
Group changes don't propagate
Security adds "High Risk Project" as a group in Okta. The AI policy that should restrict everyone in it keeps letting them through, because the policy engine never heard about the new group. Auditors find out at evidence collection time.
Your IdP feeds it. Every PortEden control consumes it.
Identity flows in from your IdPs over SCIM 2.0, OIDC, and SAML 2.0. PortEden's identity layer normalizes it once, then every downstream control — access, policy, audit, redaction — reads from the same source.
Identity sync coverage, end to end.
Six dimensions of coverage — providers, protocols, attributes, group types, lifecycle events, and mappings. The same coverage applies uniformly across every integration and every AI client downstream.
Identity providers
Which IdPs are supported?
First-class SCIM 2.0 + OIDC + SAML 2.0 connectors for the directories enterprises actually run.
- Microsoft Entra ID (Azure AD)
- Okta
- Google Workspace
- JumpCloud
- Ping Identity
- OneLogin
- Any SCIM 2.0 + OIDC compliant IdP
Protocols
Which standards does PortEden speak?
SCIM 2.0 for provisioning. OIDC for auth. SAML 2.0 for SSO. OAuth 2.1 for delegated authorization.
- SCIM 2.0 for user and group provisioning
- OIDC (OpenID Connect) for authentication
- SAML 2.0 for SSO and federation
- OAuth 2.1 for delegated authorization
- JWT-bound session tokens
- mTLS for SCIM endpoint hardening
User attributes synced
What identity data flows through?
Standard SCIM attributes plus IdP extensions and OIDC custom claims surfaced as PBAC variables.
- Email and display name
- Department and cost center
- Manager and reporting chain
- Employment status (active, suspended, terminated)
- Title, role, clearance level
- Custom attributes via SCIM extensions
- OIDC custom claims for partner-org context
Group types
Which group structures map cleanly?
Whatever your IdP already maintains — security, distribution, dynamic, nested — maps without translation.
- Security groups
- Distribution lists
- Dynamic groups (attribute-computed)
- Nested groups with full expansion
- Role-mapped groups
- Cross-tenant federated groups
Lifecycle events
Which IdP events drive policy changes?
Every SCIM lifecycle event triggers policy recomputation in real time, with reconciliation as a backstop.
- Create — new user provisioned
- Update — attribute or profile change
- Deactivate — soft removal with grace period
- Reactivate — restore prior policy bindings
- Delete — hard removal with audit retention
- Group-membership change (add, remove, move)
- Periodic reconciliation deltas
Mapping & transforms
How does identity become policy?
Group → role bundle. Attribute → policy variable. Claim filter → PBAC predicate. Multi-IdP merge for partners.
- Group → role bundle mapping
- Attribute → PBAC policy variable
- Custom OIDC claim filters
- Multi-IdP merge rules for partner orgs
- Conditional bindings (env, tenant, time)
- Versioned, change-controlled mapping edits
Connect. Sync. Bind. Enforce.
1. Connect
One-time SCIM connector and OIDC client setup in your IdP admin console. Certificate exchange. PortEden registers as a downstream app the same way any SaaS does — no agent, no elevated directory privileges.
2. Sync
Real-time push from your IdP for create, update, and deactivate events. A periodic reconciliation backstop pulls the full directory on a schedule so drift can't accumulate silently between events.
3. Bind
Map IdP groups to PortEden role bundles and policy groups. Map IdP attributes — department, clearance, manager, custom claims — to PBAC policy variables. Mappings are versioned and audited like any other policy edit.
4. Enforce
Policy decisions use the latest identity state at request time, not yesterday's snapshot. A deprovisioning that just happened is honored on the very next AI request. Every decision records the identity revision it ran against.
A group change in Entra ID, enforced in three seconds.
Here's a real lifecycle event. Sam moves from Sales to Legal in your IdP at 14:02:17. By 14:02:20, the next AI request from Sam evaluates under Legal policies. Every step is recorded.
Entra ID pushes the group-membership change to PortEden's SCIM endpoint. Payload signed, source verified, queued for evaluation.
Sales role bundle removed. Legal role bundle added. PBAC variables (department, ethical-wall side) refreshed to the new values.
Sam's next AI request runs under the new bundle. Sales-only data sources return denied or filtered. Audit trail records both the IdP event and the policy recomputation.
Every SCIM event is recorded in the audit trail with the policy version that was live before and after. An auditor asking "what could Sam see at 14:02:19?" can replay the decision against the exact policy state.
Same IdP event, two very different outcomes.
Citations, not vague reassurances.
Identity sync maps directly to the access-management clauses your auditor checks. Evidence is exportable from the audit trail in formats your auditor accepts.
Every IdP your workforce uses. Every source the AI reaches.
One identity layer, six regulated workflows.
Read the IdP, don't replace it.
PortEden does not store identity, run a parallel directory, or compete with your IdP for ownership of users. It reads from the system you already trust and binds policy to it — so deprovisioning happens upstream and we honor it instantly.
IdP-native, not a parallel directory
PortEden doesn't store identity; it reads it. Deprovisioning happens upstream in your IdP and we honor it instantly. There's no second source of truth to keep in sync, no second admin role to manage.
Group-aware, attribute-rich
PortEden uses the same groups your IdP already manages — security groups, dynamic groups, nested groups, distribution lists. No new directory to maintain, no parallel role taxonomy. Just bind policy to what already exists.
Real-time + reconciled
SCIM events drive instant changes. A periodic full reconciliation acts as a backstop so drift can't accumulate silently. If the IdP fails to send an event, the next reconciliation catches it and writes a recovery record.
Pairs well with
Identity sync questions
What is identity sync and why do I need it for AI access?
Which identity providers are supported?
How fast does a deprovisioning propagate?
Do I need to install an agent in my IdP?
What happens if my IdP is unreachable?
Can PortEden read custom attributes from my IdP?
How are nested groups handled?
What about partner orgs and federated identity?
Are sync events themselves audited?
Does this support break-glass and emergency access?
What evidence does this produce for SOC 2 CC6.2 and CC6.3 auditors?
What pricing tier includes identity sync?
Ready to make your IdP the source of truth for AI access?
Wire up SCIM and OIDC in under 15 minutes. Enterprise adds SAML 2.0 federation, multi-IdP merge, break-glass workflows, and SIEM export of identity events.