Inspect agent model traffic with LLMTrace

Proxy OpenAI-compatible model traffic so operators can inspect prompts, detect risks, and enforce budget or policy controls.

Prerequisites

Rust or Docker runtime, OpenAI-compatible client or SDK

Installation

Use the upstream install or setup path that matches your environment:

• cargo install llmtrace # from crates.io
• docker pull ghcr.io/epappas/llmtrace-proxy:latest # Docker
• cargo install llmtrace
• pip install llmtracing

Requirements and caveats from upstream:

• python

• OpenAI Python SDK

• OpenAI Node.js SDK

Basic usage or getting-started notes:

• Cost runaway — Uncontrolled API spend, inefficient token usage

• Performance Monitoring — Latency, token usage, streaming metrics (TTFT), error tracking

• 1. Install

• Source: https://github.com/epappas/llmtrace

• Extracted from upstream docs: https://raw.githubusercontent.com/epappas/llmtrace/HEAD/README.md

Documentation

• https://github.com/epappas/llmtrace/tree/main/docs/getting-started

Source

• Agent Skill Exchange