Agent Evaluation Harness and Red-Team Retainer, Monthly Ongoing on Azure AI Foundry

The first Microsoft Marketplace Azure AI Foundry-native agent evaluation and red-team operating retainer, combining the Azure AI Foundry Evaluations SDK, Microsoft AI Red Teaming Agent, Microsoft Defender for AI Services, and Microsoft Sentinel data lake correlation. A 12-month minimum monthly retainer that operates what a one-time evaluation setup builds: weekly automated regression, monthly OWASP and MITRE ATLAS red-team cycles, and quarterly deep-scenario assessments for agents already in production.

Built entirely on Microsoft Azure. This is a professional service for Azure, Kriv AI's primary product: every evaluation and red-team cycle runs against your own Azure AI Foundry agents.

What this retainer covers:

• Weekly automated regression testing via the Azure AI Foundry Evaluations SDK, with drift detection
• Monthly OWASP LLM Top 10 and MITRE ATLAS red-team attack batteries
• Quarterly deep-scenario red-teaming and jailbreak corpus refresh
• 24/7 named engineers and a quarterly board risk pack at Enterprise

Who buys. CISO or Head of AI Platform as single signer (Foundation, single entity, agents already in production). CISO and Chief AI Officer as dual signers (Standard, multi-entity, 24/7 coverage). Full C-suite with Board Risk Committee approval (Enterprise, named engineers, quarterly board pack).

Foundation — $15,000 per month, 12-month minimum, single entity, 3 agents, business-hours coverage. Weekly CI/CD-gated regression testing, monthly OWASP and MITRE ATLAS attack batteries, jailbreak corpus refresh, and a monthly executive report. Business-hours coverage, 4-hour P1 response.

Standard — $32,000 per month, 12-month minimum, 3 entities, 5 agents, 24/7 coverage. Foundation scope expanded to 5 agents and 3 entities, 24/7 P1 coverage with a 2-hour response SLA, a monthly executive briefing, custom evaluator authoring, and a full regulated-industry compliance overlay.

Enterprise — $55,000 per month, 12-month minimum, 10 entities, 10 agents, named engineers. Standard scope at production scale with named, US-based red-team engineers, a 1-hour P1 response SLA, a weekly executive readout, a quarterly board risk pack, and multi-region red-team coverage.

Add-on. +$5,000 per month per additional agent beyond tier baseline.

Frameworks. NIST AI RMF Measure and Manage functions; ISO/IEC 42001:2024 AIMS ongoing evidence; OWASP LLM Top 10 v2; MITRE ATLAS v5.4; HIPAA Audit Controls where healthcare is in scope; SR 26-2 (supersedes SR 11-7) ongoing model monitoring where financial services is in scope; SEC Item 106 and Item 1.05 for public-company issuers.

Why Kriv AI. Built on Anthropic's Claude models available in Azure AI Foundry. SAM.gov active federal contractor. US-based delivery, with named engineers at Enterprise tier. A native Azure AI Foundry service model, not a platform product, so there's no separate platform licensing fee stacked on top.

Engagement boundary. Ongoing operating retainer only, not a one-time evaluation harness setup, not an incident-response execution service (Kriv advises; the customer's own team executes), and not a HITRUST, SOC 2, or ISO 42001 certification audit. Customer retains agent ownership and remediation authority. 12-month minimum term, quarterly billing in advance, 90-day written termination notice. Claude in Azure AI Foundry is commercial-only — not available for FedRAMP-bound federal workloads.