Quick Start

Option A: OpenTelemetry (Zero Code if Already Instrumented)

Already emitting OpenTelemetry traces from your agents? Point your existing OTLP exporter at MeshAI and you are done. No SDK, no proxy, no code changes.

Get a telemetry key

Go to app.meshai.dev -> Connect -> Create telemetry key. The key needs only the telemetry:write scope.

Point your exporter at MeshAI

export OTEL_EXPORTER_OTLP_ENDPOINT=https://api.meshai.dev/api/v1/ingest
export OTEL_EXPORTER_OTLP_HEADERS="Authorization=Bearer msh_YOUR_API_KEY"
export OTEL_SERVICE_NAME=my-agent

Run your agent

MeshAI auto-discovers agents from the service.name resource attribute and extracts token usage from gen_ai.* conventions. Agents appear in the registry as telemetry arrives. See the OpenTelemetry guide for details on what gets extracted.

Option B: Proxy (Fastest, Zero Code)

The MeshAI proxy sits between your agents and LLM providers. One env var change and you’re monitoring.

Get your proxy key

Go to app.meshai.dev → Settings → Proxy Keys → Create.

Set the env var

# For Anthropic
export ANTHROPIC_BASE_URL=https://proxy.meshai.dev/v1/anthropic/k/YOUR_PROXY_KEY

# For OpenAI
export OPENAI_BASE_URL=https://proxy.meshai.dev/v1/openai/k/YOUR_PROXY_KEY

Run your agent

No code changes needed. All requests automatically flow through MeshAI.

Option C: Python SDK

Install

pip install meshai-sdk

Initialize and register

from meshai import MeshAI

client = MeshAI(api_key="msh_...", agent_name="my-agent")
client.register(
    framework="crewai",
    model_provider="openai",
    model_name="gpt-4o",
)

Enable auto-tracking

# Automatic heartbeats
client.start_heartbeat()

# Track usage (or use framework wrappers)
client.track_usage(
    model_provider="openai",
    model_name="gpt-4o",
    input_tokens=1500,
    output_tokens=800,
)

What You’ll See

After connecting, your dashboard shows:

Agent Registry: all your agents with status, framework, model
Cost Attribution: spend per agent, team, and model
Anomaly Detection: cost spikes, behavioral drift, security alerts
Governance: policy enforcement, audit trail, compliance score

Next Steps

OpenTelemetry Ingestion

OTLP endpoints, span conventions, and what gets extracted

Proxy Setup Guide

Detailed proxy configuration for all providers

Framework Guides

Auto-tracking for OpenAI, Anthropic, CrewAI, LangChain, AutoGen

Create Policies

Set up your first governance policy

API Reference

Full API documentation

Getting Started

Proxy (Zero-Code)

Framework Guides

Governance

EU AI Act Compliance

Billing

Option A: OpenTelemetry (Zero Code if Already Instrumented)

Option B: Proxy (Fastest, Zero Code)

Option C: Python SDK

What You’ll See

Next Steps

OpenTelemetry Ingestion

Proxy Setup Guide

Framework Guides

Create Policies

API Reference

​Option A: OpenTelemetry (Zero Code if Already Instrumented)

​Option B: Proxy (Fastest, Zero Code)

​Option C: Python SDK

​What You’ll See

​Next Steps

OpenTelemetry Ingestion

Proxy Setup Guide

Framework Guides

Create Policies

API Reference

Option A: OpenTelemetry (Zero Code if Already Instrumented)

Option B: Proxy (Fastest, Zero Code)

Option C: Python SDK

What You’ll See

Next Steps