"At a glance, I can see what AI at the company did"...Traceforce Reveals Use Tracking Technology
Find out how a company uses AI by asking questions in everyday language
Stop dangerous tasks before they are carried out
Enlarge image As companies start using AI agents, new security tools are being developed to track what both AI systems and employees do inside company systems.
On the 20th, Traceforce, an AI security startup in Silicon Valley, introduced a tool called "Tracegraph." It links users, AI models, Model Context Protocol (MCP), which is a standard for connecting company data and tools to AI, and token usage.
Tracegraph is a tool that shows how AI is used across a company in a single graph. It connects users, applications, and AI models. It also shows MCP connections and token usage, which help link AI to outside tools.
For example, an administrator can ask in everyday language, "Which team has doubled its token usage this month?" The tool can find the answer without the administrator having to check different dashboards separately.
The reasons companies need to manage AI use are also changing. In the past, security teams mainly focused on which systems employees and software programs were allowed to access. But AI agents can use the permissions they are given to read files or change data directly.
To address this problem, Traceforce focuses on identifying what AI agents are allowed to do and tracking the actions they actually take with those permissions. If it detects dangerous behavior, it sends a warning or blocks the action. It also stops tasks that are hard to undo, such as deleting data, before they happen.
Traceforce also checks the security of MCP, which allows AI agents to connect to outside programs. Using its open-source security tool, "MCP X-Ray," Traceforce found five security weaknesses in four types of open-source MCP servers. Each weakness was assigned a Common Vulnerabilities and Exposures (CVE) identification number. Two were rated "critical," the highest risk level.
Enlarge image Traceforce is also bringing its security technology to the U.S. business security market. On the 13th, local time, it sponsored the Dallas CISO XC Fall Conference in Texas and held a session for business leaders. On the 15th, it presented its work on AI agent security at the P1 Technologies 2026 Technology Summit in California.
Traceforce was founded last year by CEO Xia Hua and Chief Technology Officer Varun Wadwa, who previously worked at Oracle and Clumio. Y Combinator (YC) and Antler invested in the company. Traceforce also took part in YC's Summer 2026 startup demo day last month.
Traceforce is expanding its research and development team in South Korea. It opened a Korean subsidiary in Seoul in July. The Korean team handles engineering and research and development for the company's U.S. headquarters. The company is currently hiring back-end engineers.