Guides

Record, inspect, and explain a real agent run

Practical workflows for first capture, Claude Code profiling, and semantic Agent Flamegraphs.

Quick start guide

Getting started with AgentSight

Install AgentSight, record a local agent command, and export a first run artifact for inspection or sharing.

Agent guide

Record a Claude Code session and inspect the model, process, file, network, and resource paths that shaped the run.

Analysis guide

Aggregate agent sessions into semantic flamegraphs for tokens, time, operations, files, and network activity.