How a request flows through G8KEPR

Every call your AI apps, APIs, and MCP tools make can pass through G8KEPR's four protection pillars, and related events are tied together by a shared correlation ID so you can trace a request across the steps it touched. This guide walks through that path, step by step.

The four pillars

A request can touch any or all of these, depending on what it's doing:

• ▸AI/LLM protection — inspects prompts and model responses for injection, jailbreaks, and unsafe output.
• ▸API protection — checks authentication, rate limits, and schema on your API traffic.
• ▸MCP protection — governs Model Context Protocol tool calls, arguments, and results.
• ▸Agent/gateway protection — applies policy as agents chain actions or route to model providers.

The request lifecycle

1. 1.Arrival. A request reaches G8KEPR and is associated with a correlation ID that stays attached through the downstream steps.
2. 2.Identity and scope. We resolve the caller's organization, key, and tenant so checks and logging stay isolated to your account.
3. 3.Inspection. The request is evaluated against the pillars that apply. Detections that apply share the same request context, so an MCP tool call and the LLM prompt behind it can be evaluated together rather than in isolation.
4. 4.Policy decision. Findings resolve to an outcome such as allow, flag, or block, based on your configured policies and thresholds.
5. 5.Enforcement. Allowed traffic is forwarded to its destination (your model provider, API, or tool). Blocked traffic is stopped before it reaches the destination.
6. 6.Record. The outcome, the pillars involved, and the decision are written to your event log under the same correlation ID.

Following one request

Because the ID is shared, a single lookup can show the path a request took. An illustrative event record looks like this:

{
  "correlation_id": "req_8f3a2c19",
  "pillars": ["ai", "mcp"],
  "decision": "blocked",
  "reason": "prompt_injection_detected"
}

Search that correlation_id in your dashboard to review the related prompt, MCP call, detection, and enforcement action together.

Next steps

• ▸Set up your first key
• ▸Configure policies and thresholds
• ▸Read the event log