# MeshAI Labs

> MeshAI Labs is the OpenTelemetry-native Agent Control Plane: the governance,
> observability, and EU AI Act compliance layer for enterprise AI agent
> deployments. MeshAI closes the EU AI Act compliance lifecycle by triggering
> risk reassessment when agent behavior drifts (Step 11 → Step 4 of the
> compliance lifecycle described in Nannini et al., arXiv:2604.04604, 2026).

## Product

- [Home: The Agent Control Plane for Enterprise AI](https://meshai.dev): Definitional overview, four pillars, EU AI Act compliance positioning
- [AI Agent Governance Platform](https://meshai.dev/governance): Policy-as-code enforcement, audit-grade record-keeping, human-in-the-loop approvals, non-human identity management, EU AI Act compliance evidence
- [Features](https://meshai.dev/features): Agent Discovery & Registry, ML-Powered Anomaly Detection, Cost Intelligence, Governance & Compliance Engine
- [Pricing](https://meshai.dev/pricing): Starter ($299/mo), Professional ($799/mo), Enterprise ($1,999/mo)
- [Use Cases](https://meshai.dev/use-cases): Enterprise deployment scenarios across regulated industries
- [AI Agent Observability](https://meshai.dev/observability): What agent observability covers, the signals that matter, how it differs from LLM observability, instrumenting agents with OpenTelemetry
- [Deployment Models](https://meshai.dev/deployment): Multi-tenant SaaS, EU-region single-tenant on request, customer-hosted for design partners; data residency and sovereignty posture per model
- [Comparisons](https://meshai.dev/vs): Where runtime agent governance sits relative to the device, observability, and GRC layers
- [MeshAI vs Jamf AI Governance](https://meshai.dev/vs/jamf): Device-layer AI governance (installed tools on managed Apple endpoints) compared with runtime agent governance (what agents did, on any OS or cloud)
- [About MeshAI Labs](https://meshai.dev/about): Company, positioning, founder
- [Getting Started](https://meshai.dev/getting-started): Pilot partner onboarding guide
- [Free Silent Failure Scanner](https://meshai.dev/silent-failure-report): ghostspan, a free local-first scanner (pipx install ghostspan) that reads OTLP JSON or Langfuse trace exports and reports retry loops, degraded tool calls, and cost anomalies that returned 200 OK; no upload, no API key
- [Contact](https://meshai.dev/contact): Pilot program inquiries

## Guides & Blog

- [AI Agent Governance: A Complete Guide for 2026](https://meshai.dev/blog/what-is-ai-agent-governance): What agent governance includes, how it differs from bot governance and AI safety, the frameworks that apply
- [What Is an AI Agent Control Plane?](https://meshai.dev/blog/ai-agent-control-plane-explained): Definition of the Agent Control Plane category
- [Shadow AI: The Risk Isn't the Chatbot. It's the Agent.](https://meshai.dev/blog/shadow-ai-agents-risks-detection): What shadow AI is, why autonomous agents raise the stakes, and how to detect and govern unsanctioned AI
- [EU AI Act Compliance for AI Agents](https://meshai.dev/blog/eu-ai-act-compliance-ai-agents): High-risk obligations and required evidence
- [EU AI Act Article 12: Record-Keeping Requirements](https://meshai.dev/blog/eu-ai-act-article-12-record-keeping): Automatic logging obligations
- [EU AI Act Compliance Lifecycle: The Feedback Loop](https://meshai.dev/blog/eu-ai-act-compliance-lifecycle-feedback-loop): Drift to risk-reassessment edge
- [AI FinOps: Why Most of Your Token Spend Is Waste](https://meshai.dev/blog/ai-finops-token-spend-waste): Cost-per-successful-task attribution
- [The AI Agent Governance Layer](https://meshai.dev/blog/ai-agent-governance-layer-two-layer-architecture): The two-layer architecture, capability by capability, against Langfuse, Arize Phoenix, LangSmith and Datadog LLM Observability
- [AI Agent Observability vs Governance](https://meshai.dev/blog/ai-agent-observability-vs-governance): Why collecting telemetry is not governing, and where the two layers divide
- [EU AI Act Article 6: Annex III vs Annex I](https://meshai.dev/blog/eu-ai-act-article-6-2026-vs-2027): The two high-risk enforcement tracks and which one enterprise agents fall under
- [How to Monitor and Optimize AI Agent Costs](https://meshai.dev/blog/ai-agent-cost-optimization): Token-level attribution, model routing, budget guardrails, caching, anomaly detection
- [US AI Deregulation vs EU AI Act Obligations](https://meshai.dev/blog/us-ai-deregulation-eu-ai-act-still-applies): Why the June 2026 US executive order does not reduce EU obligations
- [Durable Agent Telemetry for EU AI Act Records](https://meshai.dev/blog/meshai-claude-code-durable-agent-telemetry): Why best-effort telemetry fails Article 12, and the fsync-before-anything-else design

## Technical Foundation

- **Ingest:** OpenTelemetry `gen_ai.*` semantic conventions. Vendor-neutral. Active SemConv contributor (OTel SIG issues #159, #160: gen_ai.agent.paused/.resumed/.checkpointed and gen_ai.agent.type execution-model values, filed May 2026).
- **Deploys alongside, not replacing:** Datadog, New Relic, Honeycomb, Dynatrace, Grafana Tempo, Jaeger, Arize Phoenix, Langfuse. One YAML config change in your OpenTelemetry Collector to add MeshAI as a parallel destination.
- **Regulatory grounding:** EU AI Act Articles 6 (risk classification), 12 (record-keeping), 13 (transparency), 26 (deployer accountability), 50 (GPAI transparency), 2(1)(c) (extraterritorial reach).
- **Conformance target:** prEN 18229-1 (CEN/CENELEC JTC 21, working draft Jan 2026 under Standardisation Request M/613).
- **Academic grounding:** Nannini et al., "Towards a Compliance Architecture for the EU AI Act" (arXiv:2604.04604, April 2026). MeshAI implements the Step 11 → Step 4 feedback edge from Figure 5.

## Framework Coverage

OpenAI, Anthropic, CrewAI, LangChain, LangGraph, AutoGen, Google Gemini, AWS Bedrock, LlamaIndex, Agno, Pydantic AI, Semantic Kernel. Enterprise platforms via proxy: Microsoft Copilot Studio, Salesforce Einstein, ServiceNow AI, Zapier AI.

## About

- **Founded:** 2025
- **Stage:** Pre-seed
- **Founder:** Henrique Veiga Curi
- **GitHub:** https://github.com/meshailabs-org
- **LinkedIn:** https://www.linkedin.com/company/meshai-labs