Prefactor is an observability and evaluation platform for AI agents running in production. Rather than charting failures after the fact, it scores every run live across quality, hallucination rate, drift, risk and token cost, then connects those scores to actions such as alerting, gating or a kill switch. The product drops into an existing stack through TypeScript and Python SDKs with native support for LangChain, Claude, the Vercel AI SDK, OpenClaw and LiveKit. Prefactor also maintains an open-source evaluation library and free tooling including an LLM token and cost calculator, an agent reliability grader, a prompt-injection test and an agent ROI calculator. Its documentation maps controls to compliance frameworks including the EU AI Act, GDPR, ISO 42001, NIST AI RMF, SOC 2, ISO 27001, HIPAA and the OWASP LLM Top 10, and it publishes case studies for support bots, voice customer support and internal agents across banking, legal, manufacturing, retail and telecommunications.
Funding
No funding rounds have been publicly disclosed.