Proxy Overview

• Automatic token tracking: every request’s tokens and cost recorded
• Governance enforcement: policies evaluated in real-time (<5ms overhead)
• No code changes: just change one environment variable

​

How It Works

Your Agent → MeshAI Proxy → LLM Provider (OpenAI, Anthropic, etc.)
                 ↓
          Token tracking
          Cost attribution
          Policy enforcement
          Anomaly detection

​

Supported Providers

​

Key Features

• Pass-through: Responses are identical to direct provider calls
• Streaming: Full SSE streaming support
• No content storage: MeshAI never stores prompt or completion text
• Fail-safe: If the proxy is unavailable, configure your agent to fall back to direct provider URLs