Identity Mesh — November 17, 2026
Release Date: November 17, 2026
Release Type: Mixed (New Features, Improvements, Bug Fixes, Security Fixes)
Overview
This release lets you see when a sign-in connection has stopped working and fix it yourself, finds cryptography and service accounts on the endpoints your security tools manage, lets you contact asset owners about findings, and introduces Ask Mesh:
- Identity provider health and secret replacement — see when an identity provider's credentials are failing or about to expire, and replace the client secret yourself.
- Cryptographic discovery on CrowdStrike endpoints — collect certificate and key details from the hosts CrowdStrike manages, and choose which scanner version runs.
- Service accounts from endpoint telemetry — a guided search for CrowdStrike NG-SIEM, Palo Alto Cortex XDR and Microsoft Defender for Endpoint that adds the service accounts found running on your hosts.
- Outreach — contact the owner of an asset about a finding by email or Slack, singly or as a campaign, with an approvals queue.
- Network endpoints — Qualys TLS scan results appear as network endpoints in the asset inventory, matched to their host and graded by the post-quantum and FIPS 140-2 policy templates.
- Ask Mesh — a search palette, an assistant panel and a results view reachable from every page. Ask Mesh is off until it is enabled for your tenant.
It also gives the ServiceNow CMDB connector more control over what it collects, clarifies the identity graph, fixes two defects, and updates third-party components for published vulnerabilities.
New features
Identity provider health and secret replacement
When a sign-in connection stopped working — a client secret that had expired or been rotated, for example — Mesh showed nothing wrong, and people who signed in successfully at the identity provider came back to Mesh and were refused. Settings now shows you the problem, and you can fix it.
What you get:
-
Health on the Identity providers screens. Each identity provider shows Healthy, Failing or Not checked yet, when it was last checked, and, when it is failing, since when and why. Health is checked for Microsoft Entra ID, Okta and OpenID Connect providers.
-
Client secret expiry for Microsoft Entra ID. The screen shows when the secret expires, or that it has expired. To see the date, grant Mesh the
Application.Read.Allpermission on your app registration. Expiry is not available for other identity providers. -
Replace secret. Paste a new client secret on the Identity providers screen. Mesh checks the new secret with your identity provider before saving it, and tells you in plain language if it is rejected. Sign-in uses the new secret within five minutes.
-
Axiad can replace the secret for you. If you are locked out and cannot sign in to make the change, Axiad can replace the secret on your behalf while assisting your tenant.
-
A clearer sign-in error. When sign-in fails because the identity provider rejects Mesh's credentials, the page says that your organization's identity provider settings need updating and to contact your administrator.
-
Every replacement is recorded in the audit log.
Why it matters: an expired or rotated secret used to lock everyone out of Mesh with nothing on screen to say why, and recovering took a support request. You can now see the problem, and fix it.
Cryptographic discovery on CrowdStrike endpoints
The CrowdStrike connector for endpoint discovery uses CrowdStrike Falcon Real Time Response to scan the hosts you choose for certificates and keys, so what sits on your endpoints appears beside the cryptography Mesh collects elsewhere.
What you get:
-
Certificate and key details from Windows, Linux and macOS hosts — location, format, algorithm, key size or curve, and validity.
-
Choose the scanner version for each connector. Select Automatic to follow the recommended version, Pinned to hold a specific version, or Pilot to run a new version on one Falcon host group while every other host stays on the base version. From the Scanner tab you can promote or stop a pilot and scan on demand. A change applies from the next campaign, never in the middle of one, and earlier findings keep the version that produced them.
-
Provenance on each asset. The asset's provenance panel shows the source, how it was discovered, when and on which host it was observed, and the scanner version and checksum that produced it.
-
Signed scanners. Scanner releases are signature-verified before Mesh uses them.
-
What you need. Create the connector from Manual setup under Add connector. The Falcon API client needs the scopes Real time response: WRITE, Real time response (admin): WRITE, Hosts: READ and Host groups: READ, and the Falcon response policy for the hosts in scope must allow Real Time Response, custom scripts, put and get. Contact your Axiad representative to have endpoint scanning set up for your tenant.
Why it matters: endpoints hold certificates and keys that no directory or cloud connector can see. Pinning a version lets you hold a scanner steady while you try a newer one on a few hosts, and recording the version means you can always say how a finding was produced.
Service accounts from endpoint telemetry
The CrowdStrike NG-SIEM, Palo Alto Cortex XDR and Microsoft Defender for Endpoint connectors each gain a guided Service accounts search that adds the service accounts seen running on your hosts to Mesh as non-human identities.
What you get:
-
A Service accounts search type beside the Process and DNS searches in each connector's settings. It takes no search terms; you give it a label and, if you wish, a look-back window such as 48h or 7d (default 7d).
-
Service accounts as non-human identities. Built-in and system accounts, Windows services, group managed service accounts and named service accounts are collected, each with its security identifier, the number of hosts it was seen on and when it was first and last seen.
-
Scope that varies by vendor. CrowdStrike and Cortex XDR cover Windows, Linux and macOS hosts. Defender for Endpoint covers Windows devices and keeps 30 days of advanced hunting data, so a longer look-back window finds nothing more.
Known limitations:
-
The search finds accounts seen running on hosts that have the vendor's endpoint agent. Accounts that never run anything are not found, so it is not a substitute for your directory, and an empty result does not prove an account does not exist.
-
Windows domain accounts are not merged with accounts collected from Active Directory.
Why it matters: service accounts that only ever sign in to endpoints are missing from directory and cloud connectors, but your security tools already see them.
Outreach
Outreach lets an administrator contact the owner of an asset about a finding, follow it up with reminders, and escalate it if nobody answers.
What you get:
-
Email and Slack. Add a channel under Settings, Outreach channels. For Slack, paste a bot token and select Test connection; the token must pass before the channel can be saved. Email needs no setup. If the preferred channel is not set up, Outreach falls back to email.
-
Profiles. Under Settings, Outreach channels, Profiles, define the channel, number of reminders, how long to wait before escalating, the fallback approver, and whether cases need approval before they are sent.
-
Your own wording. Write the message to the recipient and the escalation message for each profile, using placeholders such as the recipient's name, the asset and the finding. A message can be up to 2,000 bytes; if you leave it empty, Mesh uses its built-in text.
-
Send singly or as a campaign. Use the Manual outreach trigger or Bulk outreach campaign widgets. A campaign can target up to 1,000 recipients and reports its cases as pending, reminded, escalated, resolved or failed.
-
An approvals queue. When a profile requires approval, an administrator approves or rejects each case before it is sent.
Known limitations:
-
Configuring channels and profiles, sending, and approving are limited to administrators.
-
Outreach is notify-only: recipients cannot reply or act from within Slack or the email.
Why it matters: the people who own an asset are rarely the people who read the findings. Outreach carries the finding to them, and keeps asking until it is dealt with.
Network endpoints
Mesh now records Qualys TLS scan results as network endpoints — a host, a port and transport, and the TLS protocol versions and cipher suites the endpoint accepts.
What you get:
-
A Network endpoint tab in the asset inventory, listing each endpoint's host, port, last seen date and status, and a detail view with its protocols, cipher suites and linked machine.
-
Matched to the machine. Mesh matches an endpoint to a machine by its DNS name. One match links it; no match, or more than one, is shown as Unresolved or Ambiguous, and Mesh never guesses. A scan result with no DNS name is not recorded.
-
Graded for post-quantum readiness and FIPS 140-2. The post-quantum and FIPS 140-2 policy templates grade the TLS protocols each endpoint accepts, so an endpoint still accepting TLS 1.0 is reported against the protocol. A Network endpoint option appears when you choose what a policy applies to, and Network endpoints appear in the PQC readiness, compliance coverage and assets in scope widgets.
-
A path from machine to endpoint to software in the asset hierarchy.
Known limitations:
-
Qualys scan results Mesh collected before this release remain as software and cryptography entries until they are cleaned up, so an endpoint can be counted twice for a time.
-
Cipher suites are listed but not yet graded, so an endpoint's risk reflects its protocols only.
-
An endpoint's risk does not yet raise the risk of its machine.
-
Qualys does not say which software serves a port, so Linked software is empty for Qualys endpoints.
-
Network endpoints are not included in exports.
Why it matters: a TLS endpoint is a place where cryptography is in use, whether or not Mesh knows which software is behind it. Seeing it by host and port is the start of grading it.
Ask Mesh
Ask Mesh puts search and an assistant within reach of every page.
Ask Mesh is off for every tenant after this release. It is turned on per tenant, and it needs the AI assistant to be on as well. Contact your Axiad representative to have it enabled.
What you get once it is enabled:
-
A search palette (Ctrl/Cmd+K) to find a person, a non-human identity, a machine, software or a certificate by name. Results come best match first, the palette suggests questions as you type, offers Did you mean? when a name matches nothing, and lets you clear your recent searches.
-
An assistant panel beside the page you are on, which expands into a larger results view. A button on every signed-in page opens it, and it replaces the earlier AI drawer.
-
Lists open in the results view. When an answer is a list of machines, software or certificates, choose Open in results to see the same list with columns, sorting and paging, and open a certificate's detail from the table.
-
Chat history on shared workspaces. On a shared workspace, each person's widget and canvas chats are saved for that person. Before, only the owner of the workspace had chats saved, and an administrator chatting on a workspace they did not own could overwrite the owner's.
-
Continue a widget's chat in Ask Mesh. The conversation carries over without repeating the question.
-
Ask how to do something. Answers to how-to questions draw on Mesh documentation, and show the documentation pages they refer to.
-
Which fix helps most. Ask which fix clears the most risk and Ask Mesh ranks fixes by the findings they clear, the score reduction and the expected loss avoided.
-
How identities connect. Ask how a person reaches an asset, who is in a group, how two identities compare, or where an entity appears.
-
Download an answer as a report. Save an answer as a PDF or Excel file with the question, the source, the as-of time and its tables.
-
Your connected agents and tools. Ask Mesh can use the A2A agents and external MCP servers your tenant has connected, as the earlier AI drawer did.
-
Personal data stays out of the AI model's view. Names, email addresses and other personal fields that Mesh's own tools return are replaced with placeholders before they reach the AI provider. If Mesh cannot read its placeholder map, it refuses the request instead of sending it.
-
Support sessions. While Axiad is assisting your tenant, answers cannot be copied, downloaded or followed through links.
Known limitations:
-
With Ask Mesh on, the earlier AI drawer, and the conversation history in it, is not available.
-
Exporting a list from the results view to CSV includes up to 1,000 rows.
Why it matters: finding one thing among hundreds of thousands, and asking a question about what you are looking at, no longer means leaving the page.
ServiceNow CMDB: Control what is collected and how much
The native ServiceNow CMDB connector now collects only the data sets you select, and shows the limit on the records it keeps to link machines to their software.
What you get:
-
Linked records per run is a setting in the Tuning step, from 25,000 to 200,000 (default 100,000), and at least twice the rows per page. Mesh checks the value before you can move on. Raising Records per run to 200,000 used to leave software installations and certificates capped at 100,000 per run, with nothing on screen to say why.
-
Only the data sets you selected are collected. A connector whose saved configuration still carried data sets you had not ticked could keep collecting them. Mesh now applies your selection every time the connector runs.
-
Containers and Vulnerabilities show as "Not available yet" in the Choose data step, grayed out until collecting them is supported. If an existing connector already collects one, it stays selected until you clear it. All other data sets are unchanged.
Why it matters: the Tuning step now explains the limits that shape a large first load, and a connector no longer reads data you did not ask for.
Identity graph: Clearer relationships and kinds
The identity graph now describes what it draws more accurately.
What you get:
-
Supervision is labeled as supervision. A manager and the people who report to them were linked by an edge labeled can impersonate, drawn from one end only. The link now reads supervises or managed by, and shows on both people's graphs.
-
Kinds and colors agree. The identity kinds, colors and shapes in the graph now come from one definition, so a hardware asset is no longer drawn as a non-human identity in the 2D view.
Why it matters: the graph is read as a statement of privilege. A manager relationship should not read as the ability to impersonate.
Shadow AI counts include Not allowed tools
In the agentic AI inventory, the Shadow AI and Suspected shadow AI counts and filters now include tools your Runtime Agent Policies mark Not allowed.
What you get:
-
A banned tool that ran counts as Shadow AI, and one seen only by its website counts as Suspected shadow AI. Registered agents are in neither count.
-
A banned tool appears under both labels, so do not add the cards together.
Why it matters: the Shadow AI count now covers the tools you have already banned as well as the ones you have not yet decided about.
Easier filtering and a cleaner Linked Assets list
What you get:
-
Press Enter to add a filter. Typing a value in a widget filter and pressing Enter now adds the filter, as selecting Add filter does.
-
A cleaner Linked Assets list. Owner and People no longer appear among the Linked Assets in views where they cannot be opened, so the link no longer lands on an empty "no linked assets" page. Owner can still be opened from the Graph view or the Owner field.
Improvements
FAIR annualized loss expectancy (ALE) report is more consistent
-
The ALE trend uses the same method as the headline figure, so the two now agree. Expect a one-time step in the trend on the day you upgrade.
-
Tenants with no risks see "no history yet" instead of an error.
Bug fixes
-
Algorithm Posture Matrix drill-down — sorting while a Name filter was applied on a large tile timed out and showed a load error. It now shows the sorted page; on very large tiles the page can appear without a total count.
-
Support sessions — an ended Axiad-assisted session could be left without its end-of-session audit record. The end of a session and its audit record are now written together.
Security fixes
- Third-party components used by Mesh services and connectors are updated for published vulnerabilities.