What's new
Recent additions to SecureAuth Agent Authority. This page collects the current year's highlights, newest first.
October 2026
Microsoft 365 Graph covers Microsoft To Do
Microsoft 365 Graph adds 24 tools for Microsoft To Do: task lists, tasks, checklist items, linked resources, and attachments. Agents can now pull a user's open tasks into a daily briefing next to their mail and calendar. To keep To Do read-only, allow list_todo_* and get_todo_* in a policy. If you added this resource earlier, add the Tasks.ReadWrite delegated scope to your Entra app registration and grant admin consent again. Each user also needs to reconnect before the To Do tools work.
September 2026
Call the management API with an API key
You can now call the management API from your own code, such as a dashboard that reads audit events or a script that manages policies. Open Settings, go to the API keys tab, and create a key with only the permissions it needs. When you create a key, you can grant only permissions you hold for the whole organization. You can rotate or revoke a key at any time. See Authenticate with an API key.
Create custom roles
Besides the three built-in roles, you can now create your own roles from the permissions you choose and assign them to people. Assigning a role adds permissions without removing the person's existing permissions. You can't create or assign a role that grants more access than you have. See Roles.
Manage Agent Authority from your AI agents
The built-in SecureAuth Agent Authority resource gives your agents tools to call the management API. An agent can then review policies or read audit events for you. The tools available to each agent depend on its user's permissions and your gateway policies. Calls use that person's existing identity, so no separate resource sign-in is needed. See Manage Agent Authority from your AI agents.
Find risky non-human identities in Microsoft Entra
Connect a Microsoft Entra directory as a non-human identity (NHI) source, and the gateway checks it on a schedule. It reports risky configurations as findings, such as credentials that never expire or apps that accept sign-ins from outside your tenant. You can acknowledge or dismiss an open finding, then reopen it if needed. The NHI Agents tab lists the agents that the directory knows about.
SecureAuth enables this feature for each organization. You also need permission to view NHI data. Contact SecureAuth if the feature isn't available to your organization.
Match policy rules on network address and new groups
Policy conditions can now match the address a call came from. Use request.ip.inCidr() to allow or block calls by address or range, for example to allow writes only from your office network. See Network conditions. You can also type a group the gateway hasn't seen yet into a rule's scope, so the rule is ready before that team signs in. See Scope.
See which rule decided each tool
When an agent connects, its Agent Trail entry now groups tools under the rule that decided their access. It includes allowed, conditionally allowed, and denied tools. Each rule name links to the rule. Tools that no rule matched appear under No matching rule · default deny. See Event details.
Set how long resource connections last
On a connectable resource's Settings tab, you can now set a maximum age, an idle timeout, or both. These limits apply to users' connections. A background check deletes stored credentials for connections that exceed a limit. This can happen a few minutes after the limit is reached. The user must reconnect before an agent can use the resource on their behalf. See Connection lifetime.
One account for each person
The gateway matches sign-ins with the same email address to one account within your organization, regardless of how they sign in. Someone who uses a coding agent but never signs in also gets an account, based on the email address in their agent telemetry. Their usage then shows a name instead of a bare email address. A new Seen in column and filter on Human Identities shows where the gateway has observed each person. See Users.
See one person's Authority Graph
A person's page under Human Identities has a new Authority Graph tab. It shows their agents, the policies that decide their access, and the resources they reach. The graph uses their group memberships and agent instances to determine access. See One user's authority.
Download analytics as CSV files
Each tab on Enforcement Traffic (MCP Usage, LLM Usage, and Security) now has a Download button. It exports the tab's data as a zip of CSV files. The export uses your selected filters and date range. Ranked lists and session exports include up to 256 rows. The included manifest.txt file identifies any lists that exceed this limit. See Analytics.
More resource integrations
Apollo, Fireflies, Userpilot, and Vitally join the catalog. Zoom adds tools for AI search, Hub files, and My Notes. If you set up Zoom earlier, add the new scopes listed under Required scopes to your Zoom app, then ask each user to reconnect. See Resource integrations for the full catalog.
August 2026
Set rate limits on tool calls
You can now cap how many tool calls agents make. This contains a runaway agent, protects an upstream service from a flood of calls, and stops one team using up everyone's quota. Count the budget per user, agent instance, MCP server, or tool, over a window you choose. Set a rule to Log only first to see what it would catch, then switch it to refuse calls. See Rate limits. SecureAuth turns enforcement on for each organization, so contact SecureAuth if you don't see the Usage Limits page.
See AI spend by team, model, and person
The Cost page now breaks spend down by team, so you can see which teams drive your AI spend and which licensed seats nobody uses. Each team shows its model mix and how many of its seats are active. To see one person's spend, open their profile, then select the Cost tab. GitHub Copilot appears next to your other assistants in usage, cost, and session data once you send its telemetry to the gateway.
Connect an agent to one resource instead of all of them
The connect dialog now offers a scope picker, so you can give an agent only the resource it needs instead of all of them. Keep the default, All resources, or pick one resource by name. The agent then loads fewer tools and can reach nothing else. See Connect an agent.
Tools grouped by the service they come from
A resource that covers several upstream services, such as Microsoft 365, now groups its tools by service. The grouping appears on the resource details page and in the policy tool picker, so you can grant access to one service instead of the whole bundle. See Resources.
Webhooks are open to every organization
Every organization can now use webhooks. They push audit events to your own systems as they happen, so alerts reach your SIEM or chat tool without anyone polling the audit log. Built-in destinations cover Slack, Microsoft Teams, PagerDuty, CrowdStrike, Datadog, and Splunk HEC, plus any HTTPS endpoint. You create and manage them yourself, and a condition on each webhook narrows which events it delivers.
Run data protection checks on your own AI provider
Data protection can use an AI judge to spot sensitive data that no fixed pattern would catch. You can now point that judge at your own AI provider instead of a shared one. The tool responses it reads then stay with a provider you control, which is what you need when data residency rules require your data to stay in your own region. See Inference. SecureAuth turns this on for each organization, so contact SecureAuth if you don't see it.
Microsoft 365 Graph covers Teams
Microsoft 365 Graph adds tools for Teams chats, channels, and teams, so agents can read and post in Teams under your policies. If you added this resource earlier, add the 13 new delegated scopes to your Entra app registration and grant admin consent again. Each user also needs to reconnect before the Teams tools appear.
More resource integrations
More resource integrations joined the catalog, which now covers more than 30 services. Every one runs under the same policies and audit trail as the rest. See Resource integrations for the full catalog.
More agents you can connect
Gemini CLI and Antigravity join the agent catalog. See Agent integrations for every agent you can connect and its setup steps.