> ## Documentation Index
> Fetch the complete documentation index at: https://docs.meshai.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Introduction

> MeshAI is the Agent Control Plane — see, govern, and comply

# What is MeshAI?

MeshAI provides unified **observability**, **governance**, and **cost intelligence** for enterprise AI agent deployments. It discovers, monitors, and governs all AI agents across your organization — regardless of framework or vendor.

## The Problem

Organizations are deploying AI agents faster than they can track them. This creates:

* **Agent sprawl**: 144:1 non-human identities per employee, most ungoverned
* **Reliability gaps**: 20-step agent chains have only \~36% end-to-end reliability
* **Cost opacity**: Token spend is invisible and unattributed across teams
* **Compliance risk**: EU AI Act high-risk obligations provisionally postponed to December 2, 2027 (2026 Digital Omnibus); audit-ready evidence still needs 12+ months of history

## Four Pillars

<CardGroup cols={2}>
  <Card title="Agent Discovery & Registry" icon="radar">
    Auto-discover all AI agents. Detect shadow agents, maintain a living inventory, monitor health in real time.
  </Card>

  <Card title="ML-Powered Anomaly Detection" icon="triangle-exclamation">
    Four detection algorithms catch cost spikes, reliability decay, behavioral drift, and security threats every 5 minutes.
  </Card>

  <Card title="Cost Intelligence" icon="dollar-sign">
    Token-level spend attribution by team, project, and agent. Budget guardrails with automatic enforcement.
  </Card>

  <Card title="Governance & Compliance" icon="shield-halved">
    Eight policy types enforced in real-time. HITL approval workflows. [Oversight metrics](/governance/oversight-metrics) and incident evidence packs prove human oversight is real, not decorative. [Agent-to-agent delegation mapping](/governance/relationships) traces multi-agent chains. EU AI Act readiness scoring.
  </Card>
</CardGroup>

## Three Ways to Connect

<CardGroup cols={3}>
  <Card title="Proxy (Zero-Code)" icon="bolt" href="/proxy/overview">
    One env var change. All LLM requests flow through MeshAI's transparent proxy for automatic monitoring.
  </Card>

  <Card title="SDK (Code)" icon="code" href="/sdk-reference/installation">
    Python SDK with auto-tracking for [11 frameworks](/framework-guides/openai): OpenAI, Anthropic, CrewAI, LangChain, AutoGen, and more.
  </Card>

  <Card title="MCP Server" icon="plug" href="/getting-started/mcp-server">
    Governance and observability for MCP-compatible tools that don't support a custom base URL, like Claude Desktop, VS Code, and JetBrains IDEs.
  </Card>
</CardGroup>

Using **Claude Code**? See the [Claude Code connector](/getting-started/claude-code) for durable, crash-proof telemetry backed by a local write-ahead log.