Tanium Launches AgentSight to Inventory AI Agents and MCP Servers on Endpoints
Tanium introduced AgentSight, a product for discovering AI agents, local models and Model Context Protocol servers across endpoint fleets. The company’s November 17 announcement says the visibility spans Windows, macOS and Linux, including on-premises and air-gapped environments.
The initial discovery capability maps AI components to the device on which they run and the person who installed them. Tanium says it can identify coding assistants, desktop AI applications, IDE and browser extensions, local models and configured MCP servers, giving IT and security teams an inventory of tools that may otherwise operate outside formal approval.
Tanium also described capabilities to prioritize AI risk findings, record agent actions in sequence, trace them to a session and prompt, and alert when an agent deviates from its normal behavior. The company presented those controls as additions beyond the visibility available at launch.
The announcement addresses a practical governance problem: AI tools can run with an employee’s access, while security teams may not know which components are installed or what connections they use. AgentSight is intended to bring that endpoint activity into the organization’s existing inventory and response workflows.


