# AgentTrust OS

> AgentTrust OS is the enterprise governance layer for agentic AI systems. It certifies AI agents before production, enforces policy on every agent action at runtime, and keeps a full audit trail for compliance. Agent = Model + Harness; AgentTrust OS is the trust layer in the harness.

## Products

- [Trust Certify](https://agent-trust.tech/solutions/trust-certify/): Pre-production certification — security, reliability, compliance and guardrail testing against your agents before they touch production.
- [Trust Runtime](https://agent-trust.tech/solutions/trust-runtime/): Real-time enforcement at the harness layer. Validates every agent action against enforcement contracts, scores confidence and risk, auto-approves low-risk actions, escalates edge cases and blocks policy violations.
- [Trust Audit](https://agent-trust.tech/solutions/trust-audit/): Full-trace observability. Every decision, tool call, approval, escalation and block is logged for compliance reporting and executive dashboards.
- [All solutions](https://agent-trust.tech/solutions/): Overview of the product suite and challenge briefs (prompt injection defense, deterministic-first enforcement, multi-turn threat detection).

## Framework integrations

- [LangGraph](https://agent-trust.tech/integrations/langgraph/): Trust Runtime wraps graph nodes as enforcement checkpoints in stateful LangGraph pipelines.
- [CrewAI](https://agent-trust.tech/integrations/crewai/): Per-agent permission scopes and tool-call validation for multi-agent CrewAI crews.
- [Model Context Protocol (MCP)](https://agent-trust.tech/integrations/mcp/): Runtime policy governance for MCP tool servers; every tool invocation is validated before execution.

## Core concepts

- [AI Agent Harness](https://agent-trust.tech/blog/ai-agent-harness/): The infrastructure around a model — orchestration, context, state, tool access control, sandboxing, verification loops and observability.
- [Agent framework vs agent harness](https://agent-trust.tech/blog/agent-framework-vs-harness/): How to tell orchestration code from the harness a production agent needs.
- [Runtime validation gate](https://agent-trust.tech/blog/ai-agent-runtime-validation-gate/): Why runtime validation must be inline and deterministic, not an after-the-fact eval dashboard.
- [Confidence scoring](https://agent-trust.tech/blog/ai-agent-confidence-scoring/): A numeric certainty estimate for every agent decision, used to route to auto-approve, escalate or block.
- [11 scoring dimensions](https://agent-trust.tech/blog/ai-agent-scoring-11-dimensions/): The dimensions an AI agent should be scored on before production.
- [Zero-trust AI agents](https://agent-trust.tech/blog/zero-trust-ai-agents-agenttrust-runtime/): Why system prompts are not a security boundary and how Trust Runtime enforces one.
- [OWASP LLM06 Excessive Agency](https://agent-trust.tech/blog/owasp-llm06-excessive-agency/): What excessive agency is and how to prevent it.
- [AI for security vs security for AI](https://agent-trust.tech/blog/ai-security-vs-security-for-ai/): Governing what AI agents are allowed to do.
- [Agent governance gap](https://agent-trust.tech/blog/ai-agent-governance-gap/): Why existing tools leave agentic AI in production without formal governance.

## Framework guides

- [LangGraph governance in production](https://agent-trust.tech/blog/langgraph-governance-production/)
- [CrewAI safety guardrails](https://agent-trust.tech/blog/crewai-safety-guardrails/)
- [MCP runtime policy governance](https://agent-trust.tech/blog/mcp-runtime-policy-governance/)

## Compliance and audit

- [Compliance resources](https://agent-trust.tech/compliance/): ISO 42001, SOC 2 and GDPR DPIA resources for regulated industries.
- [12-field AI audit trail checklist (2026)](https://agent-trust.tech/blog/ai-audit-trail-checklist-2026/): The schema every AI-influenced decision needs to be reconstructable.
- [AI agent observability and audit trail](https://agent-trust.tech/blog/ai-agent-observability-audit-trail/): Operational, governance and regulatory observability layers.
- [GDPR DPIA for AI agents](https://agent-trust.tech/blog/gdpr-ai-agent-dpia/)
- [SOC 2 Type II for LLM deployments](https://agent-trust.tech/blog/soc2-llm-compliance/)
- [ISO 42001, EU AI Act and AIUC-1 on one control plane](https://agent-trust.tech/blog/secure-agents-iso-42001-eu-ai-act-aiuc-1/)
- [OSFI E-23 vs SR 26-2 vs EU AI Act](https://agent-trust.tech/blog/agentic-ai-regulatory-comparison-2026/)
- [Agent risk tiering for model risk management](https://agent-trust.tech/blog/agent-risk-tiering-mrm/)

## Comparisons

- [LangSmith alternative: governance, not just observability](https://agent-trust.tech/blog/langsmith-alternative-governance/)
- [AI agent certification platforms](https://agent-trust.tech/blog/ai-agent-certification-platform/)
- [AI agent trust platform for the enterprise](https://agent-trust.tech/blog/ai-agent-trust-platform-enterprise/)

## Company and resources

- [Pricing](https://agent-trust.tech/pricing/): Current plans and tiers.
- [Blog](https://agent-trust.tech/blog/): All engineering and compliance articles.
- [Documentation](https://docs.agent-trust.tech/docs/): Product and SDK documentation (TypeScript and Python SDKs).
- [LinkedIn](https://www.linkedin.com/company/agent-trust/)
- [YouTube](https://www.youtube.com/@AgentTrust-OS)