The core problem it solves is the "shadow AI" sprawl. In most offices, people are using a dozen different AI apps and plugins, and security teams have no idea which data sources these tools are actually hitting. Traceforce handles this by deploying a lightweight binary and browser extension to map out the connectivity graph between the AI app, the MCP, and the actual tool being triggered.
Deployment and Setup #
The rollout process is straightforward for IT teams:
-
Install the Go-based binary and Node JS browser extension on employee laptops or VMs.
-
The system spends about 30 minutes indexing live data.
-
All active AI agents and their connected MCPs appear on a centralized dashboard for the security team.
From a workplace adoption angle, the biggest hurdle is usually "Big Brother" syndrome. However, since content inspection happens locally on the device and prompts aren't stored by default, it's an easier sell to employees. It shifts the conversation from "we are watching you" to "we are protecting the device from leaky plugins."
Technical Breakdown #
The heavy lifting here is the mapping of attack paths. Because every AI app updates its features weekly, the developers have to manually parse configurations and logs to maintain the connectivity graph.
Tech Stack: Go (binary) and Node JS (extension).
Visibility: Detects AI apps and their specific MCP connections.
Control: Allows security admins to block predefined high-risk tool calls in real-time.
Open Source: They've released a dynamic MCP pentesting tool for vulnerability detection.
https://github.com/traceforce/mcp-xray
It's an interesting approach to AI workflow security, especially for companies trying to balance developer autonomy with strict compliance. Instead of banning tools, they're just adding a monitoring layer to see where the data is actually flowing.
Next Web Tech Meets SSH: Rendering HTML →