About MeshAI Labs
We're building the OpenTelemetry-native Agent Control Plane — because the EU AI Act requires more than your observability stack collects.
Our Mission
Enterprise AI agent adoption is accelerating — but governance isn't keeping up. CyberArk reports 45+ non-human identities per human employee in the average enterprise, most of them ungoverned. A 20-step agent chain has only ~36% end-to-end reliability. And less than 10% of companies actively govern their AI agents.
MeshAI exists to close this gap. We're the OpenTelemetry-native regulatory evidence layer above your existing observability stack — framework-agnostic, vendor-neutral, and built around the EU AI Act's actual requirements rather than as a productivity-tool retrofit. The data layer (Datadog, New Relic, Honeycomb, LangSmith) collects agent telemetry. MeshAI produces the audit artifacts — Annex III classification, Article 6(3) derogation documentation, Article 26 deployer-accountability bundles — that regulators actually demand.
Henrique Veiga Curi
Founder & CEO
Henrique brings deep experience in distributed systems and cloud infrastructure from his time at Amazon Web Services (AWS). Having seen firsthand how fragmented ecosystems can limit innovation, he founded MeshAI to solve the governance and observability challenge in the rapidly growing AI agent space.
With a background in building scalable infrastructure that powers millions of applications, Henrique recognized that enterprises are deploying AI agents faster than they can track them. Without a control plane, agent sprawl becomes a security, cost, and compliance liability. MeshAI is his vision for giving organizations full visibility and control over their AI agent ecosystem.
Why Now
Organizations must demonstrate audit trails and governance for AI systems.
CyberArk: non-human identities vastly outnumber employees — most are ungoverned.
A 20-step agent chain has barely a third chance of completing successfully.
The vast majority of organizations have no active AI agent governance.