SPICEDB + MICROSOFT ENTRA ID
SpiceDB: A Relationship-Based Access Control (ReBAC) Solution That Works With Microsoft Entra ID
Microsoft Entra ID decides who someone is and which applications they can sign in to. SpiceDB decides what they can do inside those applications, down to every document, project and record. Together they carry your identity strategy all the way into the product.
For most organizations, Entra ID is the foundation of identity: single sign-on to thousands of applications, strong authentication, Conditional Access and governance over who belongs to which group. It answers the most important question about every session, and it answers it well.
Inside the applications you build, a different question takes over. Who can see this document? Who can edit this project, approve this invoice or open this customer account? The answer depends on relationships that live in the product rather than the directory: who owns what, what was shared with whom, and which folder, workspace or account an object belongs to. That is the question SpiceDB was built to answer, and it plugs straight into Entra ID.
What Microsoft Entra ID does well
- Single sign-on everywhere. Standard OpenID Connect, OAuth 2.0 and SAML sign-in for your own applications and thousands of others.
- Strong, adaptive authentication. Multi-factor and passwordless sign-in, with Conditional Access policies that weigh user, device, location and risk before a token is issued.
- Users and groups as the source of truth. One directory for employees, partners and guests, with groups, app roles and app assignments.
- Governance built in. Access reviews, entitlement management and lifecycle workflows keep group memberships and app access accurate as people join, move and leave.
- Identities for software too. Workload identities let applications, services and automation authenticate the same way people do.
- A rich API. Microsoft Graph exposes the directory, including change tracking, so other systems can stay in sync.
If your priority is controlling who can sign in to which application, under what conditions, Entra ID is the right control point.
Where application-level questions begin
Entra ID can place a user's app roles and group memberships in the tokens it issues, and that works well for coarse decisions such as "is this person an administrator of this app?" Object-level decisions are different, for three reasons:
- The data isn't in the directory. Ownership, shares, folders, projects and customer accounts live in your product. No directory should hold millions of them.
- Tokens have size limits. Microsoft documents that a JWT carries at most 200 group memberships (150 for SAML). Beyond that, the token contains an overage indicator instead, and the application has to call Microsoft Graph to find the user's groups.
- Lists need a different answer. A search page or an AI assistant needs to know every object a person can reach, not just whether one request is allowed.
The pattern that scales: let Entra ID own identity and group membership, and let a permissions database own how those people and groups relate to everything in your product.
What SpiceDB adds
SpiceDB is an open-source permissions database inspired by Google's Zanzibar, the system Google described for authorizing products such as Drive, Calendar and YouTube. It stores relationships, such as this group can edit this project or this folder contains this document, and computes permissions from them on request.
- Object-level permissions at very large scale. In published load tests, SpiceDB served one million permission checks per second against one hundred billion relationships with a 5.76 ms P95. Read how.
- Groups and hierarchies, natively. Nested groups, organizations, workspaces, folders and individual shares, in one model. Grant access to an Entra group once and it flows to every member.
- Answers for lists, not just checks. Lookups return everything a person can access, for search, dashboards and AI retrieval.
- Consistency you control. ZedTokens let an application require answers that reflect the latest change, so someone removed from a group doesn't see new content.
- A schema developers own. Permissions live in a short, readable schema that can be tested in CI and explored in a browser playground.
- Runs where you do. Open source under Apache 2.0, self-hosted on Azure or anywhere else, or fully managed by AuthZed, including private connectivity over Azure Private Link.
Two systems, two questions
| Microsoft Entra ID | SpiceDB | |
|---|---|---|
| The question | Who is this, and may they sign in to this application under these conditions? | Can this person act on this object? And which objects can they reach? |
| What it stores | Users, groups, app roles, app assignments and credentials | Relationships between people, groups and your product's objects |
| Shines at | Single sign-on, strong authentication, Conditional Access, lifecycle and access governance | Ownership, sharing, hierarchies, multi-tenant products, permission-aware lists, search and AI retrieval |
| Where decisions happen | At sign-in and token issuance, and through app assignment | Inside your applications and data platforms, through check and lookup APIs |
| Who usually owns it | Identity and security teams, in the Entra admin center | Application and platform engineers, through schema as code |
How they work together
Everything here uses standard protocols and APIs, so nothing in your Entra tenant has to change. These are the patterns teams use most:
1. Entra ID signs people in, SpiceDB uses their object ID
Your application signs users in with Entra ID as usual and reads the user's object ID (the oid claim) from the token. That identifier becomes the subject of every SpiceDB check. Use it rather than email or user principal name, which can change. Multi-tenant applications can also carry the tenant ID (tid) to keep each customer's directory separate.
2. Entra groups become SpiceDB relationships
Keep Entra ID as the source of truth for groups, and mirror memberships into SpiceDB, either by having Entra provision users and groups to your application over SCIM or by following directory changes through Microsoft Graph. Nested groups map directly onto SpiceDB's model, and because memberships live in SpiceDB rather than in each token, group size limits stop being a concern.
definition user {} // id: the Entra object ID (oid)
definition group { // synced from Entra ID
relation member: user | group#member
}
definition project {
relation owner: user
relation editor: user | group#member
relation viewer: user | group#member
permission edit = owner + editor
permission view = edit + viewer
}
Grant a group access once, for example project:apollo#editor@group:engineering#member, and everyone in that group, including members of nested groups, can edit the project. Remove someone from the group in Entra ID and the change flows through to every object the group could reach.
3. Layered decisions: Entra at the door, SpiceDB inside
Conditional Access, app assignments and app roles decide who gets into each application and under what conditions. Inside the application, SpiceDB decides what each person can do with each object. A request has to pass both, which gives you defense in depth without duplicating rules.
4. Governance that reaches every object
Entra ID Governance keeps memberships accurate through access reviews, entitlement management and lifecycle workflows. Because those memberships flow into SpiceDB, a review that removes someone from a group removes everything that group granted, across every application that uses SpiceDB. SpiceDB's lookups can also show reviewers exactly what a group gives access to.
5. Applications and AI agents with the right boundaries
Services and AI agents authenticate with Entra workload identities, the same way people do. SpiceDB then decides which documents and records they can reach, and an agent acting for a person can be limited to exactly what that person could see, including filtering retrieval before anything reaches the model.
When to add SpiceDB to your Entra environment
- Your applications have sharing, collaboration, teams, folders or workspaces.
- You rely on group or role claims in tokens and are running into size limits or overly broad roles.
- You serve many customers, each with their own Entra tenant, users and roles.
- Lists, search and reports must show each person exactly what they can access.
- AI assistants or agents work over documents with different owners and audiences.
- Permission logic is still scattered across services, even with Entra ID in front of them.
Getting started
- Model one application. Describe its objects and access rules as a SpiceDB schema, and try it in the browser playground.
- Map identities. Use the Entra object ID as the SpiceDB user, and choose SCIM provisioning or Microsoft Graph to sync groups.
- Add checks and lookups. Call SpiceDB wherever the application shows or changes data, starting with the screens that matter most.
- Choose how to run it. Self-host the open-source release on Azure or anywhere else, or let AuthZed run it as a managed service.
See SpiceDB alongside your Entra ID setup
Send us your company name and we’ll build a custom permission schema and a live, clickable demo of how relationship-based authorization would work in your application, designed to sit alongside your Entra tenant.
Get a custom demo · Explore SpiceDB demos · Open the Playground
AuthZed and Microsoft are independent companies. This article describes common ways customers use the two together, based on standards such as OpenID Connect, SCIM and the Microsoft Graph API; it does not describe a joint product. Microsoft, Microsoft Entra and Azure are trademarks of the Microsoft group of companies. Entra ID details reflect Microsoft's published documentation as of November 2026; some capabilities, such as SCIM provisioning and ID Governance, depend on the customer's Entra licence.