LLM-side Protection

Protect model-side execution before runtime drift becomes production risk.

Use this path when your main concern is the runtime behavior around the model: routing, fallback, guard behavior, confidence signals, approval boundaries, and governed execution before live handoff continues.

Start LLM-side Runtime Path Explore LLM-side APIs

This page keeps the LLM-side path separate from User-side evidence and Development Platform work, so customers do not get trapped in a nested integration maze.

Runtime path guard: Starting this path creates a Builder/runtime entry first. A Builder project does not mean production runtime control is complete, a first runtime proof does not enable AgentOps or Approval Gateway, and Runtime Continuation, AgentOps, or Approval Gateway requests are intent signals only until separately activated.

Core Runtime API

Route, inspect, and govern model-side runtime behavior with visibility into fallback, guard, confidence, and execution signals.

Approval / Governed Execution

Move sensitive continuation points through approval, block, escalation, and commit-boundary control before execution proceeds.

AgentOps

Use operational surfaces for runs, review, triage, queue handling, and human operator workflows around runtime decisions.

When to choose this side

Choose LLM-side Protection when execution control is the first buying reason.

This side is for teams that need production AI control before output reaches the user-facing evidence and review layer. If your first concern is behavior after the answer is produced, use User-side Protection instead. The Builder path starts access, proof, and request capture; it does not by itself complete production runtime control or activate AgentOps / Approval Gateway.

Continue to LLM-side Builder Compare User-side Protection