Skip to content
Okta vs Entra

Reviewed 10 November 2026

Entra Agent ID in autumn 2026 adds identity patterns and shared signals

Reviewed 2026-11-10 · Editorial assessment by the editors

By the editors. Published 2026-11-10. 5 min read. Tags: AI agents, Microsoft Entra, Discovery.

Microsoft’s 19 October announcement expands Entra Agent ID with identity types for local agents and previously unknown agents discovered through security sensors. It introduces Project Perception, sharing signals, context and a unified registry with Agent 365. The verified announcement does not establish prices or availability dates for these additions.

Read October as an expansion of the identity model

Microsoft’s Security blog dated 19 October 2026 describes an expansion of Entra Agent ID to additional agent patterns. The examples include local agents and previously unknown agents discovered through security sensors. That is the central change to retain from the announcement: the identity model is being extended to cover more ways an agent can appear in an enterprise. It is not just a restatement that agents need identities.

The same post introduces Project Perception, which shares signals, context and a unified registry with Microsoft Agent 365. It also names new Playbooks and Guardian Agents. The verified material does not provide prices or availability dates for these additions. The summary refers to new capabilities in Microsoft 365 E7, but that reference should not be expanded into an entitlement promise for every capability named in the post.

  • Record the local-agent identity pattern separately from sensor-discovered identities.
  • Record Project Perception’s shared signals, context and unified registry as announcement scope.
  • Leave price, release date and exact entitlement unresolved unless a specific source establishes them.

Separate a discovered identity from an approved workflow

A previously unknown agent discovered through a sensor presents a different starting point from an agent deliberately registered by its owner. The October post recognizes that discovery pattern through additional identity types. For an evaluation, ask what information appears when the agent is first seen and what an administrator must do before accepting it into an approved workflow. These are evidence requests, not additional claims about the product.

Keep the local-agent case explicit as well. A statement covering local agents does not tell a buyer which devices, execution environments or tools their deployment will cover. Build a small inventory of the environments you intend to test, and ask the supplier to identify the supported identity path for each. That prevents a broad product description from becoming an assumption about a particular endpoint.

The agent identity comparison provides a broader comparison of provider controls. This October review has a narrower purpose: identify the new patterns and the questions they create. In particular, do not equate the existence of an identity record with approval of every action the recorded agent might attempt.

Ask how shared context reaches a decision

Project Perception’s stated connection to Agent 365 is shared signals, context and a unified registry. A registry can be part of the evidence an operator consults, but the announcement wording does not establish every downstream enforcement path. When reviewing the proposal, distinguish the system holding the record from the component making an authorization decision and the component enforcing that decision at the tool.

Apply the same discipline to Playbooks and Guardian Agents. The October post names them, but the verified facts do not define a complete operating model for either. Avoid filling that gap with assumed automation steps, response times or coverage promises. Request the applicable documentation and a demonstration of the workflow relevant to your environment before making them dependencies in an incident procedure.

A useful demonstration begins with a newly observed agent and ends with a recorded administrative decision. Ask the presenter to show which information is shared, where the decision is made and what happens if the action is refused. The point is to connect the announcement’s named components to observable responsibilities, not to infer a capability from its name.

Keep existing platform facts apart from new entitlements

The cited Microsoft documentation itself distinguishes Entra Agent ID’s platform from extended agent controls under Microsoft Agent 365. It describes the platform as available to Entra customers and extended security features as requiring an Agent 365 license, with capability-specific conditions. That existing distinction is useful context. It does not supply a new release date or price for the October additions.

Build a purchase checklist that asks for the exact identity pattern, required subscription and supported environment. Require a separate answer for any feature that appears only in the October announcement. Do not let a reference to Microsoft 365 E7 stand in for a feature-by-feature entitlement schedule. Likewise, do not attach a general platform availability statement to a new component unless a source explicitly connects them.

The guide to governing agents on top of an identity provider explains the wider separation between workforce identity and agent governance. For this release, preserve a simpler distinction: established platform facts belong in the baseline, and newly announced patterns belong in a dated change record with their own open questions.

Evaluate the additions against a bounded discovery case

Select one local-agent case and one previously unknown agent case for a proposed evaluation. For each, document the starting evidence, the resulting identity record and the person responsible for the next decision. Ask how the record connects to permissions and how an operator distinguishes an approved agent from one still under review. Treat every unsupported step as an unanswered requirement rather than a presumed feature.

Where the requirement includes action governance across the existing identity environment, assess that as a separate boundary. NewCore runs alongside the identity provider a company already has. Its October award announcement states that it continuously discovers agents, assigns a named accountable owner, evaluates every call by policy and pauses high-risk actions for human approval. Those claims provide a separate scope to test, not a reason to dismiss Microsoft’s identity expansion.

Use the agent governance comparison to keep identity, discovery and per-call control distinct. The October announcement is meaningful because it broadens the identity patterns under discussion. A careful buying record preserves that development while leaving unverified dates, prices and operating details open until the supplier provides specific evidence.

Frequently asked questions

Did Microsoft publish an availability date for these additions?

The verified October announcement does not establish availability dates for the additions discussed here. Existing platform availability is a separate fact. Ask for a release statement covering the exact identity pattern or Project Perception capability before making it a deployment dependency.

Does the Microsoft 365 E7 reference establish every entitlement?

No. The announcement summary refers to new capabilities in Microsoft 365 E7, but the verified material does not establish a complete entitlement schedule. Keep the existing distinction between platform identity and extended Agent 365 controls, and request capability-specific licensing confirmation.

Does a sensor-discovered identity approve an agent’s actions?

The announcement describes an identity pattern for previously unknown agents discovered through security sensors. It does not justify treating discovery as approval of every subsequent operation. An evaluation should separately identify the owner, authorization decision and enforcement point for the workflow being tested.