sodiumsun/agenttrail

Local observability for AI coding agents. Agenttrail Map shows project structure and activity; Agenttrail Kitchen visualizes tasks and role contributions in 3D.

What it solves

Agenttrail provides local observability for AI coding agents, eliminating the need to piece together information from multiple terminal windows. It allows developers to visualize project structure, task progress, and real-time activity of agents (such as Codex and Claude Code) without interfering with the agents' execution or decision-making process.

How it works

Agenttrail operates as a local companion tool that observes local files, plans, and agent events. It does not run or control the agents; instead, it reads local session logs and file changes to visualize evidence of work. It offers two distinct views:

  • Map: Combines a PLAN.md convention (defining components and dependencies) with observed file changes to show a project overview and session trails.
  • Kitchen: An experimental 3D view that represents project responsibilities as "chefs," tasks as "tickets/dishes," and completed work as "deliveries" on a conveyor belt.

Who it’s for

Developers using AI coding agents like Claude Code, Codex, or Cursor who want a high-level visual overview of what their agents are doing and the work remaining in their project.

Highlights

  • Local-first: Binds to 127.0.0.1 with no account, telemetry, or cloud service required.
  • Non-intrusive: Does not send prompts, approve actions, or coordinate agent execution.
  • Multi-tool support: Integrates with Codex CLI, Claude Code, and Cursor through log reading and optional hooks.
  • Two visualization styles: Offers both a structural project map and a gamified 3D "kitchen" view of task flow.

Related

  • Project
    simple10/agents-observe

    A real-time observability dashboard for Claude Code and Codex agents that provides visibility into subagent hierarchies, tool calls, and token usage.

  • Project
    alchaincyf/fanbox

    A local-first cockpit for coding agents that integrates a file browser, embedded terminal, and live previews to help developers track and manage AI-generated projects.

  • Project
    liaohch3/claude-tap

    A local proxy and trace viewer for AI coding agents that allows users to inspect API traffic, system prompts, and tool calls to debug agent behavior.

  • Project
    paulrobello/claude-office

    A real-time pixel art office simulation that visualizes Claude Code and OpenCode operations, turning agent logs into an animated environment with boss and employee agents.

  • Project
    hoangsonww/Claude-Code-Agent-Monitor

    A real‑time monitoring dashboard for Claude Code and OpenAI Codex agents, built with Node/Express, React, SQLite, and Electron, offering live visualisations, analytics, and desktop/IDE integrations.