Evaluate, observe, and improve your AI agents
We give platform, ML, and engineering teams the means to measure how every agent performs in production, and make it better, with the runtime controls to enforce it.
Prefactor is an agent evaluation platform for teams building AI agents: it observes every agent run from development to production, evaluates whether the agent did its job, and shows how to improve it.
What is Prefactor?
Prefactor is an agent evaluation platform. It observes every AI agent run from development to production, evaluates whether the agent did its job, and shows how to improve it, with runtime controls to act on what it finds. It is built for platform, ML and engineering teams running agents for customers or staff, and it works across frameworks through TypeScript and Python SDKs or OpenTelemetry. Prefactor Pty Ltd is an Australian company, and its primary infrastructure runs in Australia.
Built for the teams who have to say yes
Every enterprise wants AI agents. Security teams want to say yes. But right now, there's no way to see what agents are doing, measure whether they're doing it well and improve them, or prove compliance when the auditor asks.
Prefactor operates at the agent runtime layer. Lightweight SDKs. Full observability. Enforceable policy. No proxies, no network-level hacks, just clean integration that ships with your agents.
We built Prefactor so that every stakeholder, security, compliance, product, engineering, can share a single source of truth for what agents are doing and how well they perform, and move forward together.
The gap between agent demos and agent production is evaluation
The agent landscape is fragmented. Claude, Copilot, LangChain, N8N, custom builds, every team picks their own stack. Nobody has the full picture. CISOs are dealing with shadow agents the same way they dealt with shadow IT and shadow cloud before it.
Meanwhile, governance processes designed for traditional software take nine to twelve months to approve an agent. By that point, the model is three generations old and the business has moved on. We founded Prefactor because the industry needed a single way to observe, evaluate and improve agents in production, framework-agnostic, with enforcement at the agent runtime layer.
Visibility first
You can't secure what you can't see. Visibility is the foundation, before a single policy, know what's running, who owns it, and what it's doing.
Enforce, don't suggest
Controls at the runtime layer. Not guardrails agents can reason around. Policies that hold whether the agent cooperates or not.
Framework agnostic
One plane for Claude, Copilot, LangChain, N8N, and whatever ships next. We don't pick winners, we observe, evaluate and improve them all.
Ship, don't block
The goal is production, not permanent POC. We compress approval timelines from months to days by giving every stakeholder what they need.
The people building Prefactor
Backed By
Help define how enterprises ship agents safely
We're building the agent-quality layer (observe, evaluate, improve) the ecosystem is missing. If you care about infrastructure, security, and shipping things that matter, we want to talk.