Identity fabric and identity orchestration are sold, and often bought, as if they were the same thing, and the confusion leads to programmes that buy an integration layer and expect it to fix governance, or design a grand architecture that never gets wired together. An identity fabric is an architecture: the set of loosely coupled identity services, data and interfaces an organisation runs to manage every kind of identity. Identity orchestration is one capability inside that architecture: the layer that coordinates journeys, policies, attributes and signals across those services at design time or at runtime. This guide compares the two for IAM architects, identity programme leads and security reviewers: what each is for, where orchestration sits in a fabric, how each fails, how far each reaches into non-human identities, and when consolidating onto fewer platforms beats composing many. For how to design an orchestration layer itself see the Identity Orchestration Guide; this page does not repeat it.
Key takeaways
- A fabric is the whole loom, orchestration is one shuttle. KuppingerCole describes the identity fabric as a logical infrastructure for enterprise identity and access management; orchestration is a component that makes the services in it work together.
- Neither is a product category you can buy whole. KuppingerCole states there is no standard identity fabric, and orchestration platforms coordinate services they do not replace.
- The risks differ in shape. A fabric fails through gaps and inconsistency between services; an orchestration layer fails through concentration, because it holds the secrets, keys and policy that every connected service trusts.
- Non-human identities expose the difference. Orchestration mostly acts in sign-in and request paths, while most workloads, API keys and service accounts never pass through a journey; fabric-level inventory and ownership is what covers them.
- Standards keep both replaceable. OIDC and SAML for federation, SCIM for provisioning, OAuth token exchange for delegation, the Shared Signals Framework for events and the AuthZEN Authorization API for decisions let you swap parts without rebuilding the architecture.
The short answer
Identity fabric: an architectural paradigm. KuppingerCole's definition is a loosely coupled set of identity services, typically built on one or two strong core platforms for identity governance and access management, plus complementary components, exposed through a comprehensive set of APIs and serving employees, partners, customers, devices and things, and services. The firm is explicit that a fabric is assembled from interoperable building blocks rather than bought as one suite. You do not deploy a fabric; you design toward one and measure how close you are.
Identity orchestration: a coordinating layer. It sits between applications and identity services and decides which service to call, in which order, with which data. It covers sign-in and registration journeys, attribute aggregation at token issuance, policy translation and distribution, and the movement of security signals. Policy orchestration projects such as Hexa describe the design-time form: discover native policies, translate them to and from a common policy language (IDQL), and distribute them to identity providers, clouds and data platforms. Hexa's own site frames discovery as analysing the organisation's identity fabric, which shows the relationship: orchestration operates on a fabric.
Side-by-side comparison
| Dimension | Identity fabric | Identity orchestration |
|---|---|---|
| What it is | Target architecture for all identity services and data | Integration and coordination layer within that architecture |
| Primary question answered | Which services, data sources and interfaces do we need, and how do they connect? | How does a request, journey or policy flow across the services we already have? |
| Scope | Governance, access management, privileged access, directories, identity data, NHI management, ITDR, CIAM | Journeys, routing, claims, policy translation, signal distribution |
| Identity types | All: workforce, partners, customers, devices, workloads, agents | Strongest for interactive sign-in and request paths; partial for workloads |
| Time horizon | Multi-year programme and roadmap | Project-sized, often first used for a migration or acquisition |
| Owned by | Identity architecture and programme leadership | IAM engineering, platform team |
| Success looks like | Every identity has an authoritative source, an owner, governed access and monitoring | Applications integrate once; IdP, MFA or policy engines change without application rewrites |
| Typical failure | A diagram that is never implemented; gaps between services | A new single point of failure holding every client secret and signing key |
| Key standards | SCIM, OIDC, SAML, SSF, AuthZEN, plus inventory and governance practice | OIDC, SAML, OAuth token exchange (RFC 8693), SSF and CAEP, AuthZEN, policy languages |
Where orchestration sits in a fabric
A practical way to reason about a fabric is as layers. Orchestration touches several of them but owns none, which is why it cannot stand in for the fabric.
| Fabric layer | What lives there | Orchestration's role |
|---|---|---|
| Identity data | Authoritative sources, correlation, attribute quality, NHI registry | Consumes attributes at runtime; should not become a new store of truth |
| Governance | Joiner, mover, leaver, access requests, certification, ownership | Rarely involved; may trigger workflows from signals |
| Access management | IdPs, federation, MFA, session management | Routes and sequences sign-in journeys across several IdPs and authenticators |
| Authorisation | Policy decision points, entitlements in each application and cloud | Design-time: translates and distributes policy. Runtime: calls PDPs through an API such as AuthZEN |
| Privileged and workload access | PAM, secrets managers, workload identity federation | Limited; may broker token exchange for services and agents |
| Detection and response | ITDR, posture management, logging | Distributes and acts on signals, for example session revocation on a CAEP event |
The identity data layer is where most fabrics are weakest. See the Identity Data Quality Guide for authoritative sources and correlation, and the Identity Convergence Guide for the related question of merging IGA, access management and PAM.
Threat model: the identity fabric
Architecture in words: many services, each with its own administrators, logs and data, connected through APIs and provisioning flows. The attack surface is the sum of the services plus every trust relationship between them.
- Seams between services. A leaver disabled in the HR-driven directory but still active in a SaaS application provisioned by hand; a cloud role granted outside governance. Attackers favour the service that was left out, typically through T1078.004 Cloud Accounts.
- Inconsistent assurance. One application federates through a phishing-resistant IdP while another keeps local passwords. The fabric is only as strong as its weakest entry point.
- Uncorrelated identities. Accounts that link to no person or owner cannot be reviewed or revoked. This is chronic for service accounts, API keys and OAuth applications.
- Federation trust abuse. Every service that trusts an IdP's signing key inherits its compromise. In the SolarWinds Golden SAML campaign, attackers used a stolen token-signing certificate to forge SAML tokens (T1606.002) and added credentials to existing service principals (T1098.001).
- Visibility gaps. Logs spread across services make cross-service attack paths hard to see without an identity-centric detection layer. See the IVIP Guide.
Controls: an authoritative source and owner for every identity and critical attribute, provisioning by standard (SCIM) rather than by hand, a uniform minimum assurance for entry points, signing keys in hardware with rotation, and identity posture and threat detection that spans services. See the ISPM Guide.
Threat model: the orchestration layer
Architecture in words: a component that acts as a relying party to upstream IdPs and as an identity provider to downstream applications, holds client secrets and signing keys for both sides, calls risk and verification services, and may push policy into many targets with administrative credentials.
- Concentration of secrets. Compromise of the layer, or of its administrators, yields the ability to mint assertions trusted by every downstream application and to push policy everywhere. It is a tier 0 asset.
- Assurance downgrade. NIST SP 800-63C-4 treats an orchestration broker as a federation proxy: all requirements on IdPs and RPs apply to it in each role, its assertions name it as issuer, and the FAL of the downstream connection is the lowest along the entire path, which the proxy must represent accurately. A broker that hides an FAL1 upstream link behind a stronger downstream link misleads every relying party.
- Token passthrough and impersonation. Forwarding upstream tokens to downstream services, or exchanging them with impersonation semantics, removes the record of who acted. RFC 8693 distinguishes impersonation, where A is B within the token's rights, from delegation, where A keeps its own identity and the act claim records the actor.
- Policy push credentials. Design-time orchestration needs write access to native policy in each cloud and platform. Those credentials are non-human identities with very wide reach and are easy to forget in reviews.
- Availability. A runtime layer in every sign-in path is a single point of failure; an outage stops all access, and emergency bypasses tend to become permanent.
Controls: treat the layer, its admins and its push credentials as tier 0; keep keys in hardware; preserve upstream assurance in downstream assertions; use delegation semantics with the act claim rather than impersonation; design for regional failover with a tested break-glass path. See the Identity Provider and SSO Security Guide.
NHI coverage: where each reaches
The clearest practical difference shows up with non-human identities. Workloads, pipelines and agents rarely pass through an interactive journey, so an orchestration layer sees them only when they request tokens through it.
| Identity type | What the fabric must provide | What orchestration can add |
|---|---|---|
| Service accounts in directories | Inventory, owner, purpose, credential rotation, review | Little; they authenticate directly to the directory |
| API keys and static secrets | Discovery, vaulting, rotation, revocation | Can replace some with short-lived tokens via token exchange |
| Cloud workload identities | Federation trust policy per cloud, least-privilege roles, CIEM | Can translate a common policy into native cloud policy (design time) |
| OAuth applications and SaaS integrations | Consent governance, scope review, owner | Can enforce consent and scope policy where it brokers the grant |
| AI agents and MCP servers | Agent registry, owner, delegated authority, monitoring | Audience-bound, delegated tokens for each hop; runtime policy calls |
| Devices and things | Device identity, certificate lifecycle, posture | Device signals as inputs to journeys and access decisions |
Agent traffic is where orchestration helps most with NHIs. The MCP authorization specification (revision 2026-07-28) makes an MCP server an OAuth 2.1 resource server, requires clients to send an RFC 8707 resource parameter naming the server, requires servers to validate that tokens were issued for them, and states that servers must not accept or transit any other tokens. An orchestration layer that issues per-server, audience-bound tokens through delegation fits that model; one that passes a user's token through to tools does not. See the MCP Security Guide and the OAuth Token Exchange and Delegation Guide. Ownership and inventory for everything else remain fabric work: see the NHI Ownership Guide.
Consolidate or compose: decision guidance
The fabric question is not whether to have one, since every organisation already runs several identity services; it is whether to reduce their number or add a layer that connects them.
| Situation | Lean to consolidation | Lean to orchestration |
|---|---|---|
| Number of IdPs and directories | Two or three with overlapping function | Many, after acquisitions, with no near-term retirement possible |
| Migration under way | Target platform agreed and funded | Gradual move where applications cannot all be repointed at once |
| Customer identity | One brand, one CIAM platform | Several brands or regions, multiple verification and fraud vendors |
| Team size | Small team that cannot run another tier 0 component | Platform team able to engineer for availability |
| Policy estate | Few clouds, policy owned in each | Many clouds and data platforms needing one policy intent |
| NHI problem | Inventory and ownership are the gap | Agents and services need delegated tokens across domains |
Small organisations: consolidate first. One IdP federated to everything, provisioning by SCIM, and a single governance tool give most of the fabric's value without an orchestration layer to secure. Large and federated organisations: both apply. Define the fabric target, then use orchestration to bridge the services that will not be consolidated within the planning horizon, with an explicit exit plan for each bridged service. Customer identity: orchestration is common because journeys combine verification, fraud and consent services; see the Customer IAM (CIAM) Guide.
Trade-offs: consolidation reduces seams and admin surfaces but increases dependence on one platform and makes migrations big-bang. Orchestration reduces lock-in and allows gradual change but adds a highly privileged component and a latency hop, and it can postpone consolidation indefinitely if it has no retirement plan for what it bridges.
Worked scenario
A group with 30,000 employees has two workforce IdPs after an acquisition, a legacy on-premises access manager for 200 web applications, three clouds, a CIAM platform plus a separate identity verification service, and a growing agent programme. The board has asked for "an identity fabric".
- Define the fabric target, not a product. The architecture team lists required services by layer, names the authoritative source for every identity type, including a registry for service accounts and agents, and records which services will be retired.
- Measure the gaps. Correlation shows 18 percent of cloud principals and most API keys have no owner. That is fabric work: inventory, ownership and governance, not orchestration.
- Use orchestration where it bridges. A runtime orchestration layer fronts both workforce IdPs and the legacy access manager, so applications move to one OIDC integration while IdP consolidation proceeds over 18 months. Downstream assertions carry the upstream assurance level, per SP 800-63C-4.
- Policy at design time. A policy orchestration tool expresses access intent once and pushes native policy to the three clouds; its push credentials go into PAM with approval and short-lived access.
- Agents. The orchestration layer acts as the token exchange point for agents, issuing audience-bound, delegated tokens per MCP server, while the agent registry stays in the fabric's data layer.
- Signals. The ITDR tool transmits session-revoked events through the Shared Signals Framework; the orchestration layer and SaaS applications act as receivers.
- Exit plan. Each bridged IdP has a retirement date; the orchestration layer's scope shrinks as consolidation completes.
How to implement
- Baseline: catalogue identity services, identity types, authoritative sources and trust relationships. Mark each signing key, client secret and admin credential with its tier.
- Target architecture: agree the fabric layers and which services fill each, with retirements dated.
- Close data and governance gaps first: correlation, ownership and provisioning give value regardless of orchestration.
- Pilot orchestration on one bounded journey: one application set or one acquired IdP, in shadow or low-risk mode, measuring latency and failure behaviour.
- Harden before scale: tier 0 controls, keys in hardware, delegation semantics, failover and a break-glass journey tested with rollback to direct IdP integration.
- Expand by journey, retire by service: every added integration should have a matching retirement or consolidation milestone.
Detection and measurement
- Correlation coverage: share of accounts in each service linked to a known person, workload or agent with an owner. The core fabric metric.
- Standard integration share: share of applications integrated by OIDC, SAML and SCIM rather than custom connectors or local accounts.
- Assurance fidelity: share of downstream assertions that carry the upstream authentication and federation assurance accurately.
- Orchestration blast radius: number of signing keys, client secrets and policy push credentials held by the layer, with rotation age. The trend should fall as consolidation proceeds.
- Revocation latency: time from a high-risk signal to session termination across services.
- Orchestration availability: sign-in success rate and failover test results per quarter.
Monitor the orchestration layer's admin actions, key and secret changes, new upstream or downstream trust configurations and policy pushes as tier 0 events. Alert on new credentials added to existing applications, the pattern behind T1098.001.
Common failure patterns
- Buying "a fabric": a platform is purchased under the fabric label, and the ownership, data and governance work that defines a fabric never starts.
- Orchestration as a permanent crutch: the bridge layer grows to front every service, consolidation stops, and the organisation ends up with all the old services plus a new tier 0 component.
- Assurance laundering: a broker presents a weak upstream login as strong to downstream applications, defeating their step-up rules.
- Forgotten push credentials: policy orchestration credentials with write access to every cloud sit outside PAM and access review.
- Human-only scope: the fabric programme covers workforce and customers but leaves service accounts, API keys and agents to individual teams.
- Token passthrough for agents: user tokens forwarded to tools and MCP servers, removing accountability and breaking audience binding.
Framework mapping
| Framework | Reference | Fabric | Orchestration |
|---|---|---|---|
| NIST CSF 2.0 | PR.AA-01, PR.AA-03, PR.AA-05, ID.AM-08, PR.IR-03 | Identity and credential management, lifecycle across services | Authentication routing, policy enforcement, resilience of the layer |
| NIST SP 800-53 Rev. 5 | AC-2, AC-3, IA-8, SC-12, CM-8 | Account management, component inventory, external identities | Access enforcement, key management for signing and client secrets |
| NIST SP 800-63C-4 | Section 3.3.3 proxied federation | Federation design across IdPs | Broker obligations and lowest-FAL rule |
| MITRE ATT&CK | T1078.004, T1606.002, T1098.001, T1528 | Unmanaged cloud accounts and app credentials at the seams | Forged assertions and stolen tokens if the layer is compromised |
| ISO/IEC 27001:2022 Annex A | A.5.15 Access control; A.5.16 Identity management; A.5.17 Authentication information; A.8.24 Use of cryptography | Identity management across the estate | Access control enforcement and key protection |
Practitioner checklist
- Write down the fabric target by layer, with an authoritative source and owner for every identity type, including NHIs and agents.
- Measure correlation coverage before choosing any orchestration tool.
- Decide consolidate or compose per service, with a retirement date for each bridged service.
- Classify the orchestration layer, its administrators and its policy push credentials as tier 0.
- Keep signing keys in hardware and track rotation age for every key and client secret the layer holds.
- Preserve upstream assurance in downstream assertions, following SP 800-63C-4 proxy rules.
- Use delegation with the act claim, and audience-bound tokens per MCP server, rather than token passthrough or impersonation.
- Prefer standard interfaces (OIDC, SAML, SCIM, SSF, AuthZEN) so any component can be replaced.
- Test orchestration failover and a break-glass path to direct IdP integration.
Standards and references
- NIST: SP 800-63C-4 Federation and Assertions
- IETF: RFC 8693 OAuth 2.0 Token Exchange
- OpenID Foundation: Authorization API 1.0 (AuthZEN)
- OpenID Foundation: Shared Signals Framework 1.0
- Model Context Protocol: Authorization (revision 2026-07-28)
- NIST: SP 800-207 Zero Trust Architecture
- KuppingerCole: Identity Fabrics
- Hexa: Policy Orchestration and IDQL
Related NHI Mgmt Group resources: Identity Orchestration Guide · Identity Orchestration Buyer's Guide · Multi-Cloud Identity Guide · IAM and IGA Basics