Products & Tools

Tanium launches AgentSight to track and govern AI tools on endpoints

Tanium's AgentSight, launched November 18, 2026, discovers AI agents, local models and MCP servers on Windows, macOS and Linux endpoints, mapping each to its device and installer.

Tanium launches AgentSight to govern AI tools on endpoints
Tanium launches AgentSight to govern AI tools on endpoints27707 / pixabay
By Elena Vasquez5 min read

Why it matters

  • Tanium launched AgentSight on November 18, 2026 to monitor and govern AI software on endpoints.
  • AgentSight works across Windows, macOS and Linux and maps AI components to devices and installers.
  • CTO Harman Kaur said the product extends Tanium's endpoint behavioral baselines to AI agents.
  • AI cost tracking comes through Tanium Asset; a Shadow AI page in Atlas shows daily inventory changes.
  • Detect-only mode ships first; blocking and removal of unsanctioned AI tools are in development.

Tanium launched AgentSight on November 18, 2026, a product built to monitor and govern artificial intelligence software running on endpoint devices across corporate fleets. The release extends the company's endpoint management platform to a category of software that most IT departments never formally approved: AI agents, local models and Model Context Protocol servers sitting on employee laptops and desktops.

AgentSight works across Windows, macOS and Linux. It identifies local models, coding agents, desktop AI applications, browser extensions, integrated development environment extensions and related tools. Each AI component it discovers is mapped to the device it runs on and to the person who installed it. That visibility is delivered through the Tanium console, including in on-premises and air-gapped environments — a detail that matters for regulated industries and government customers who cannot rely on cloud-hosted analytics.

Why does this launch matter now?

The release lands as companies confront a surge of AI software running outside formal IT approval processes. Many of these tools now operate on endpoints using employee access credentials. That combination — capable autonomous software acting with a human's permissions, but without oversight — has become a genuine governance problem for security and IT teams.

An AI agent that misuses its access doesn't necessarily trip traditional alarms. It uses legitimate credentials through legitimate channels. Tanium's bet is that endpoint telemetry, the discipline it has sold for years, is the right lens for this new class of software.

Harman Kaur, Chief Technology Officer at Tanium, said the product applies the same monitoring approach the company built for endpoints to software agents. "An agent misusing its access may look like legitimate activity, so you need to know what's expected and normal," Kaur said. "Tanium already builds behavioral baselines for endpoints, and AgentSight extends that discipline to agents."

The core insight is baseline-driven: rather than flagging every action an agent takes, AgentSight watches what each agent normally does and raises an alert only when behavior departs from that pattern. It's the same logic behind anomaly detection in endpoint security, pointed at a new target.

What does AgentSight actually surface?

Kaur outlined the specific information the product is intended to give customers. "It will rank AI risk findings and give a plain reason for each," she said. "It will also record what each agent did, in order, to the session and the prompt that started it, and alert when an agent departs from its own normal."

That last element — tying an ordered record of an agent's actions back to the session and the originating prompt — is significant for incident response. Teams investigating a suspicious action can reconstruct the causal chain: what instruction started the agent, what it did step by step, and which identity it acted on behalf of. Kaur framed the scope in estate-wide terms: "Endpoints, agents, the infrastructure they depend on, and the identities behind them are managed as one estate on one platform."

The product keeps a tamper-evident record of what an agent did and which identity it acted for, which gives auditors and investigators a trail designed to resist alteration.

How does it handle AI spending?

AgentSight includes AI cost tracking through Tanium's Asset product. That gives IT and finance teams a view of spending on AI software and where that software has been deployed. It's a tacit acknowledgment that unmanaged AI isn't only a security exposure — it's a line item that finance teams often can't see.

Tanium is also showing a Shadow AI page in Atlas, its platform interface. The page presents an inventory of AI software across the organization and changes since the previous day, so teams can spot newly installed tools quickly rather than waiting for periodic audits. Users can query that information in natural language and ask Atlas to explain a finding. The intent is to let teams review discovered AI tools, activity and risk in a single interface instead of stitching together exports from multiple tools.

What control features come with the visibility?

Visibility is only half the picture. Alongside discovery, Tanium is developing broader control functions that would let teams set AI policy across device fleets and then act on systems that breach those policies. The stated examples include removing or blocking unsanctioned AI tools. Tanium plans a detect-only mode first — a cautious rollout that lets customers inventory and observe before they start enforcing.

Some response measures are already available through Tanium Threat Response. They include:

  • Process termination
  • Device quarantine
  • Live Response

That means organizations that spot a rogue agent today don't have to wait for the policy engine to ship. They can act through the existing response tooling, then rely on the tamper-evident record to document what happened and who was responsible.

Where does AgentSight sit in Tanium's portfolio?

AgentSight launches as a standalone product within Tanium's wider portfolio. That portfolio also includes:

  • Security Operations
  • Endpoint Management
  • Exposure Management
  • Atlas

The standalone positioning signals how Tanium sees the market: AI governance on endpoints is emerging as a distinct budget line, not merely a feature bolted onto endpoint management. Security buyers who already run Tanium can correlate agent activity with endpoint and identity data in one place, while new customers can adopt AgentSight without the full stack.

What's the bigger picture?

The launch reflects a shift in how enterprises will have to think about AI risk. The first wave of corporate AI policy focused on approved platforms and data-handling rules. The second wave — the one AgentSight targets — is about the long tail of AI software employees install themselves, from coding agents in IDEs to browser extensions with model access to local models running entirely on-device.

Those tools inherit employee credentials, act with partial autonomy, and leave sparse logs in places security teams don't traditionally look. Mapping each one to a device and a person, recording what it did and why, and pricing the sprawl is the minimum viable governance layer most enterprises currently lack. With detect-only mode shipping first and enforcement functions still in development, Tanium is betting that customers will want to measure the problem before they try to shut it down — and that within a year, governing AI agents on endpoints will be as standard a security practice as patching the endpoints themselves.

Source: itbrief.com.au

Share this article:

More from Elena Vasquez

Elena Vasquez

Show full bio

Market editor covering technology and product news in artificial intelligence and machine learning at AI In Context.

708 articles

Related articles

  1. Tanium launches AgentSight to track AI agents on corporate devices
  2. AI Agents Move Onto Windows PCs, and Permissions Become the New Battleground
  3. Partnership on AI: Agent Frameworks Can't Show What Agents Did
  4. 83% of Firms Run Unapproved AI Coding Agents, BlueFlag Research Finds
  5. Virtana’s AI agent war room targets cross-system IT incidents

« Previous articleNext article »