ProofAgent is the accountability platform for production AI agents. It turns agent risk into deployment evidence through adversarial multi-juror scoring, production log audits, artifact reviews, signed readiness reports, and human review. The platform is built around the open-source ProofAgent Harness.

How do I test my AI agent with ProofAgent?

Install the open-source harness with 'pip install proofagent-harness', wrap your agent in a function returning AgentResponse, then call Harness().evaluate(my_agent, role, goal, knowledge, context). The harness runs adversarial multi-turn sessions and returns a /10 readiness score with traceable findings and fix recommendations.

What is adversarial multi-juror scoring?

Adversarial multi-juror scoring is ProofAgent's evaluation approach: a planner picks domain traps, a conductor applies sustained pressure across 25+ turns, and three independent juror agents score every behavior change. No single LLM call ever decides the verdict — the jury agents reach consensus or debate to a final score.

Is ProofAgent SOC 2 / HIPAA / GDPR compliant?

ProofAgent is SOC 2 Type II aligned, HIPAA-ready (BAAs available for enterprise customers), and follows GDPR best practices. Enterprise customers can deploy on-premises or in a private cloud with SSO/SAML, RBAC, tamper-evident audit logs, TLS 1.2+ in transit, and AES-256 at rest.

Can I use my own LLM with ProofAgent?

Yes. ProofAgent is BYO Harness LLM — the harness internals can run on any LLM provider (OpenAI, Anthropic, Google, local models). You bring your own model and API key; the harness orchestrates the multi-juror evaluation around it.

What metrics does ProofAgent measure?

11+ production metrics including Task Success, Hallucination Control, Safety, Policy Compliance, Memory Stability, Tone and Empathy, Manipulation Resistance, Tool Picking, Reasoning Quality, Relevance, and Drift Detection. Every metric is anchored to per-turn transcript evidence.

What is the difference between ProofAgent Platform and ProofAgent Harness OSS?

ProofAgent Harness OSS is the open-source multi-turn adversarial testing engine — Tier 1 of the platform, available standalone for developers and CI under Apache 2.0. ProofAgent Platform is the enterprise product that adds the other four tiers (production log audit, artifact review, multi-agent orchestration scoring, expert human review), a hosted dashboard, REST API, governance features, signed readiness reports, and dedicated support.

ProofAgent Harness: Adversarial Evaluation for Production AI Agents

Name: ProofAgent Platform
Brand: ProofAgent
Availability: InStock

ProofAgent Harness (arXiv:2605.24134) by Dr. Fouad Bousetouane is the open source framework for adversarial AI agent evaluation. It runs a full pipeline of planning, adversarial conducting, multi juror scoring, debate consensus, and signed reporting, and returns a readiness verdict you can ship to security and audit teams.

Single answers pass; real agents fail under pressure

Most evaluation scores a single answer with one model acting as judge. Production agents fail differently: three turns in, under pressure, through domain specific failure modes. The Harness selects domain relevant traps, sustains adversarial pressure across a full conversation, scores the whole trajectory with a panel of jurors, and produces evidence linked findings rather than one opaque number.

What the paper covers

The full pipeline: domain aware planner, adversarial conductor, three juror personas, Delphi or debate consensus, signed reporter
A 183 trap library across 11 attack families, including composite attack chains spanning five to seven turns
A six metric rubric: task success, instruction following, hallucination resistance, tool use, safety, manipulation resistance
Asymmetric evaluation: a small local model serving as the juror against frontier class agents
Multi juror consensus that measurably reduces the bias of a single model acting as judge

Why it matters for enterprises

Capability is no longer the bottleneck. The risk that keeps agents out of production is behavior under pressure: a support agent that leaks data when pushed, a finance agent that breaks policy when flattered, a tool call that fires when it should refuse. Run the Harness as a gate in CI and CD, block deployment on a failed critical metric, and ship with a signed readiness report your auditors accept.

Headline finding

Production grade agents built on GPT 5.5 and Claude Opus 4.7 fail the Harness with serious safety and manipulation resistance issues under composite chain pressure. Frontier LLMs alone are not enough. Every result is reproducible from the open source code under Apache 2.0. The companion paradigm paper, Human-on-the-Bridge (arXiv:2606.16871), formalizes the approach.

Plain language explainer on Medium: ProofAgent Harness: The Open Source Harness for Complete AI Agent Evaluation.