Entelligence-AI/agentplay

What it solves

Coding agents often perform complex tasks in the background, leaving behind opaque logs that are difficult to interpret. This tool makes those hidden session transcripts readable, allowing you to see exactly what an agent read, edited, where it failed, and what it ultimately produced.

How it works

It scans your machine for existing agent session files (such as those from Claude Code, Cursor, or Copilot), processes them into a standardized, redacted map format, and launches a local web viewer. The entire process happens locally on your machine without requiring API keys or cloud uploads.

Who it’s for

It is for developers who use AI coding agents and want to debug agent behavior, understand why specific commands failed, or audit what an agent touched within their codebase.

Highlights

• Comprehensive Visualization: Provides a radial map of every file, tool, and subagent touched, plus a playback timeline for session analysis.
• Privacy-First: Operates entirely offline; no session data ever leaves your machine.
• Broad Compatibility: Supports a wide range of agents including Claude Code, Cursor, GitHub Copilot, Gemini CLI, and more.
• Actionable Insights: Helps you identify repeated command failures, context window limits, and the specific files an agent ignored or edited.
• Zero Dependencies: Runs as a standalone tool with no complex installation requirements.