Atested Launches Cryptographically Verified Governance Platform for AI Agents
Key Takeaways
- ▸Atested provides real-time visibility and governance for AI agent actions through an HTTP proxy that intercepts and classifies every action before execution
- ▸Uses cryptographic signing (Ed25519) and hash-chaining to create independently verifiable, tamper-evident audit records of all decisions
- ▸Operates with zero performance overhead and zero token use, remaining invisible to the AI model while maintaining full enforcement capability
Summary
Atested has announced an HTTP proxy platform designed to provide real-time transparency and governance over AI agent actions before execution. The tool sits between AI models (including Claude Code and other autonomous agents) and system access, monitoring and recording every action proposed by these agents. Each action is classified, cryptographically signed with Ed25519, and hash-chained to create tamper-evident audit records that can be independently verified.
The platform addresses a critical gap in AI safety: the lack of visibility into what autonomous agents are actually doing when given access to production systems. Atested operates transparently to the AI model with claimed zero performance overhead, returning standard error responses when actions are denied so the model's capability is fully preserved. According to the materials, setup takes less than five minutes.
Atested offers tiered functionality: free operation for individual developers running AI agents locally, governance and compliance features for multi-developer teams, and enterprise-grade role-based access control with contractual SLA compliance for large-scale AI operations. The platform is designed to scale from individual use to production deployments across organizations, with an initial user base showing 259 denied actions across 5 users over a 61-day period.
- Offers free tier for individual developers and enterprise tiers with role-based access control, compliance reporting, and contractual SLAs
- Directly addresses growing concerns about AI agent safety and governance as autonomous systems gain access to production systems and critical infrastructure
Editorial Opinion
Atested represents a pragmatic shift toward operationalizable AI safety. By placing governance at the execution layer rather than trying to modify model behavior, it achieves transparency and auditability without compromising capability—a crucial trade-off as AI agents become more autonomous. The cryptographic verification approach is particularly notable: it allows independent verification without requiring trust in Atested's infrastructure, which is exactly what organizations need for compliance-critical deployments. As AI agents move from research into production systems, tools that combine simple deployment with strong cryptographic guarantees could become essential infrastructure.
