The Privent SDK gives you direct access to ACARS scoring and APE transformation as a library. No orchestration framework required. Works with any Python or JavaScript agent: custom LLM loops, internal toolchains, proprietary pipelines.
Install and initialize
Add the Privent SDK as a dependency. Initialize a client with your API key and deployment group. The SDK resolves your organization's active policy at startup.
Wrap your payload before any external call
Call `privent.scan(payload)` before sending data to any external LLM or API. The SDK runs ACARS scoring on the full payload context and returns a risk decision in under 50ms.
Apply transformation and continue
If risk exceeds threshold, APE returns a transformed payload with sensitive fields masked or fragmented. Your agent sends the safe version downstream. Every decision is logged to your Privent audit trail.
Add Privent to your existing Custom SDK pipeline without changing your architecture. One addition. Full graph-state visibility.
from privent import PriventClientprivent = PriventClient(api_key="your_api_key",deployment_group="engineering")def run_agent(user_input: str):payload = build_prompt(user_input) # your existing logicresult = privent.scan(payload)if result.decision == "BLOCK":return result.safe_response # APE-transformedreturn call_llm(result.payload) # clean payload
External gateways receive only the final prompt string. Privent reads everything in context, in real time.
Protect proprietary agent architectures built on raw LLM calls without switching to a supported orchestration framework.
Run per-tenant policy evaluation on agent payloads before they reach external models. Separate risk thresholds per deployment group.
Require all internal teams to route LLM-bound data through the Privent SDK as a policy gate, enforced at the library layer.
We integrate in under 30 minutes. No orchestration changes required. Your pipelines keep running. Privent keeps watching.
