Employees now run AI agents on their own laptops, and most security teams can’t say where those agents are or who owns them. Amplifier’s October release closes that gap. It finds the AI agents on employee devices and puts a named owner behind each one. It tags the findings your tools already produce so the right ones reach the right person faster, and gives your team one view of every employee’s devices and the gaps on them. We’re also opening a private preview for the Amplifier MCP server.

Endpoint AI Agent Discovery

See your employees’ endpoint AI agents.

The problem. Employees install and run AI agents on their laptops: productivity agents (Hermes Agent, NemoClaw, ChatGPT Work), AI browsers and browser agents (Gemini in Chrome, Copilot Mode in Edge, Comet), and coding agents (Claude Code, Cursor, Codex). Anthropic and OpenAI now give admins telemetry for their tools, but only for enterprise accounts and only for their own harness. Everything else runs outside an admin console: personal accounts, open-source agents, and the local MCP servers, skill files, and LM Studio models those agents load. Each agent acts with the employee’s credentials and sessions, so its actions look like the employee’s. You can’t govern what you haven’t found.

What Amplifier now does. Amplifier finds the AI agents on every managed device and puts a named owner behind each one. Its own discovery utility deploys through your MDM or UEM tool, inventories agents installed on employee devices, and merges them with what your connected tools already report, including Anthropic and OpenAI telemetry, EDR, identity, and others. Amplifier already maps each device to its owner, so every agent inquiry starts with someone to ask. Ampy, the AI Security Engineer, messages the employee in Slack or Teams.

Ampy confirms ownership with the employee in Slack, then asks what the agent is for.

Once the employee confirms, they state what the agent is for in their own words. Every intent on record comes from a person who vouched for the agent, so policy decisions trace back to someone accountable. If an employee says the agent isn’t theirs or doesn’t reply, Amplifier follows the rules you have set, such as removing the agent, and escalates quietly. You set the policy up front, and Ampy handles the follow-through, so nothing new lands in your queue.

The Agents view: every AI agent on employee devices, with its kind and its owner.

Why it matters. You can’t govern an AI agent you can’t see, and you can’t remediate one nobody owns. Discovered agents flow into Agent Intent & Posture Management (AIPM), a capability of the Agentic Workforce Security Platform, now in private preview. Your security team decides what happens next and can scope an agent’s access down instead of revoking it.

Finding Tags: From “What Is This” to “Who Fixes It”

Amplifier now automatically tags the findings it ingests from your existing security tools. Your team sees the pattern behind the issues and gets from issue to action faster. Tags come two ways. Custom tags follow rules your team defines in the Amplifier admin portal. Dynamic tags apply automatically to vulnerability findings, with no rules to build. The first three cover the affected product, the CVE, and whether the vulnerability is in the Cybersecurity and Infrastructure Security Agency (CISA) Known Exploited Vulnerabilities (KEV) catalog. Your team can view, sort, and search findings by tag in Security Hub and in Ampy Risk Advisor. That makes it faster to decide what reaches an employee, what goes to IT/Security, and what stays internal.

Amplifier Engagements can also now filter on these tags: engage only on findings with a given tag, without it, or with a set of tags, no matter which tool reported the finding. For example, your team can set every KEV-tagged finding to reach the device owner right away, at critical priority.

Tag filters in an engagement decide which findings reach the device owner.

A Revamped User Asset Security View

We’ve shipped a redesigned User Asset Security view in Endpoint Security Posture Management (ESPM). The new view starts with the person: each employee mapped to every device they own, with each device’s security posture and any gaps in endpoint security tools, such as EDR software that’s missing or not reporting. Summary cards open a filtered view in one click, and every column sorts and filters, so finding the stragglers doesn’t take a spreadsheet.

Why it matters. An unattributed laptop is an unprotected person, and EDR software that isn’t reporting is a control that might be silently failing. This is the view where both show up first.

The User Asset Security view: every employee, their devices, and the gaps in their endpoint security tools.

Amplifier MCP Server

The Amplifier MCP server connects your security team’s AI tools to your Amplifier data through the Model Context Protocol (MCP), the open standard AI tools use to reach outside data. Point any AI tool that supports MCP at Amplifier’s MCP server and ask a question in plain language, such as “Which devices with a KEV-tagged finding have no confirmed owner?”

Amplifier MCP is in Private Preview. Request a demo to see it, and we’ll follow up.

Asking an AI tool about Amplifier data through the MCP server.

Why This Release Matters

Your employees’ AI agents get owners. Your findings get auto-tagged to speed routing. Your team gets one view of every person’s devices and what’s missing on them. Less chasing, and the work keeps moving.

Preview access to AIPM and the Amplifier MCP server starts with a demo.
See the October release in a 20-minute demo.