A profile runs in one of two modes, and the mode determines where the platform sits relative to your AI traffic. In Enforce mode it runs in-line, directly in the path between your AI tool and the LLM. In Listen mode it runs in parallel, evaluating a copy of the traffic without sitting in its path. A profile is always in exactly one of these two modes.

The practical difference follows from that position. Running in-line, a profile can act on traffic. Running in parallel, it can only observe it.

Enforce

In Enforce mode, the platform sits between your AI tool and the LLM. Prompts and other requests are routed to it first, evaluated for signals and against the profile's settings, and acted on according to your policies. Based on those policies, the content is blocked, modified, or allowed to continue to the LLM. The LLM's response makes the return trip the same way, routed back through the platform, evaluated, and acted on before it reaches your tool.

Because Enforce sits in the path of the traffic, it can stop or change content before it reaches its destination. The signal evaluation results and the full log are visible in Monitoring.

Listen

In Listen mode, prompts and responses are forwarded to the platform for evaluation separately from the traffic going to the LLM. The platform is not in the path between your tool and the LLM, so it does not act on the content. It classifies the content, records the signal results, and passes the resulting telemetry back to your tool along with a recommended action based on your settings.

This mode is useful when you want a record of all traffic and its signal telemetry but do not need policy enforced in-line. You see what the profile would have done without it changing the traffic.

For working with traffic through the API, see the developer documentation.