Tools that make a coding-agent session visible — prompts, tool calls, subagents, file changes and whole trajectories. Mostly Claude Code, now also Codex, Cursor, OpenCode, Gemini CLI and cross-agent tools. From cinematic replays and pixel-art companions to dashboards, log viewers and observability. 32 tools, grouped.
The closest to "Gource for Claude Code" — prompts, tool calls, subagents and file changes as spatial graphs or animated replays. Highest cinematic value.
Real-time node graph of agent orchestration
Watch Claude Code (and Codex) sessions render as a live interactive node graph — see how the agent breaks down problems, which tools it reaches for, and how subagents branch and coordinate. Streams events via hooks with near-zero latency.
Prompts, subagents & file changes as an interactive graph
A VS Code extension that visualizes a Claude Code session as an interactive graph — each node is a prompt, branches show subagent spawns, and tool usage is tracked so you can see exactly which tools were invoked and how often.
Cozy 3D browser replay of a Claude Code session
Community Three.js experiments render a Claude Code session as a navigable 3D world of prompts, tool calls and file nodes in the browser. This is the closest match to the "3D graph visualizer" idea — more art installation than dashboard.
The original animated file-tree visualization
The tool that started the aesthetic: Gource renders a Git repo as an animated, growing file tree with contributors orbiting the files they touch. It accepts a custom log format, so Claude Code tool events can be converted into commit-like events and replayed the same way.
Playful, ambient viz: your session dramatized as pixel-art characters, offices and dungeons. More vibe than data, but great for making agent activity felt.
Pixel-art RPG companion for your session
A Paul Robertson-inspired pixel-art companion (Go + Raylib) that watches your session JSONL: file reads cast spells, bash swings a sword, errors deal damage, and a mana bar drains as your 200k context fills. Level up and unlock hats as you use Claude Code.
Your terminals as characters in a pixel office
Every Claude Code terminal becomes an animated character in a customizable virtual office — typing when writing code, reading when searching, waiting when it needs you. Task-tool subagents spawn as separate characters linked to their parent, with token/context health bars.
Real-time pixel-art office simulation
A pixel-art office that visualizes Claude Code operations in real time: a boss & employee agent setup, multi-floor building navigation, and visual state indicators for working, delegating and waiting.
Sessions as pixel-art dungeon heroes
A React + Canvas + WebSocket web app that turns each Claude Code session into an animated knight roaming a Sanctuary, Dungeon, Boss Arena, Merchant Shop and Tavern. Browser-based, no editor extension required.
Themed animated characters per session
A simple real-time visualizer that renders active Claude Code sessions as animated characters in themed environments — each agent walks to its workspace and raises a tool matching its current action. Ships with several swappable themes (office space, age of agents, minions, and more).
OpenCode sessions as blobs in a live office
An ambient companion for OpenCode: each running session becomes an animated blob character in a virtual office whose behaviour reflects live status across eight states, with subagents orbiting as smaller blobs.
Narrative dashboards and usage analytics — sessions, tools, costs, context, rhythm. Good for patterns over time rather than a spatial replay.
Your usage as a scroll-driven narrative
A single-page, scroll-driven story built from your local Claude Code usage data — sessions, projects, tools, coding rhythm and AI-generated insights rendered as an interactive narrative rather than a static dashboard.
Real-time dashboard with a Kanban board & buddy
A real-time monitoring dashboard (SQLite, Node/Express, React, WebSockets) that tracks sessions, agent activity, tool usage and subagent orchestration — with live analytics, a Kanban status board, notifications and a cute buddy. Ships as a web UI and native macOS/Windows app.
Full web client with tool & subagent panels
A full-featured web-based Claude Code client for managing projects, with real-time conversation-log viewing, a panel listing tool invocations with filters, and a dedicated section for agent sub-sessions. Watch sessions live from the CLI.
Local-first usage dashboard for Codex CLI
Ingests Codex CLI rollout logs into a local dashboard of calls, threads and insights — token accounting, cache diagnostics, weekly-credit projections and a "needs attention" ranking of the runs worth investigating.
Turn the raw JSONL transcripts (Claude Code, Codex, Cursor, …) into readable, browsable, shareable conversations with expandable tool calls.
JSONL session viewer — desktop, web & TUI
Browse Claude Code sessions from ~/.claude/projects: renders JSONL as readable conversations with expandable tool calls, token counts and timestamps, plus live tailing of active sessions. Runs as a desktop app, a web app, or a TUI.
JSONL → readable HTML / Markdown (CLI)
A Python CLI that converts Claude Code transcript JSONL files into clean, chronological HTML or Markdown — for single files or the whole ~/.claude/projects hierarchy with a linked index — plus an interactive TUI for browsing sessions.
Review .jsonl sessions in a clean HTML UI
A lightweight tool that renders Claude Code .jsonl files as a nicely formatted, self-contained HTML UI for reviewing what happened in a run — messages, tool calls and the files a session touched.
Session log viewer for OpenAI Codex CLI
A native desktop + web viewer for Codex CLI's rollout JSONL logs — browse conversations, tool calls and token counts, follow orchestrator/worker collaboration chains, and live-tail active sessions over SSE.
Browse, search & export your Cursor chat history
A local web UI over Cursor's SQLite state that surfaces your entire AI chat and Composer history — browse, search, export and share past agent conversations that are otherwise locked inside the editor's storage.
Offline desktop viewer with analytics
A cross-platform Tauri desktop app to browse and search agent conversation history offline, with an analytics dashboard and token stats. Beyond Claude Code it also reads Cline, Roo Code and Gemini CLI history.
Not Claude-only: one tool that reads sessions from many agents at once — Claude Code, Codex, Cursor, OpenCode, Gemini CLI, Aider, Cline. Unified session browsers, replays and cross-agent usage graphs.
Token usage across 40+ agents, as a 3D contributions graph
A CLI + dashboard that reads local session logs from 40+ coding agents (Claude Code, Codex, OpenCode, Gemini CLI, Cline, Aider, and more) and renders your activity as a GitHub-style contributions grid — flat 2D or an isometric 3D city — plus leaderboards and a "Wrapped" recap.
Local-first session search & analytics across agents
Indexes local sessions from Claude Code, Codex, Cursor, OpenCode, Gemini CLI and Aider into one searchable app, with dashboards for activity heatmaps, tool usage, velocity and per-project breakdowns — a concurrency-timeline view of everything your agents did.
Local-first macOS browser for every agent's history
A polished native macOS app to browse, search, analyze and resume coding-agent session history across Codex, Claude Code, OpenCode, Cursor Agent and Copilot CLI — transcript inspection, an image browser and a quota meter, all offline.
Agent sessions → self-contained HTML replays
Turns a session from Claude Code, Cursor, Codex, Gemini CLI or OpenCode into a single embeddable HTML file that replays the run turn by turn — prompts, reasoning, tool calls and outputs — with no server needed. Despite the name it's genuinely multi-agent.
Your agents as pixel characters in a shared office
A desktop app that turns running coding agents into animated pixel-art characters in a virtual office — Codex, Claude Code, Gemini, Cursor, OpenCode, Aider and Cline all share the room. Session search and transcript browsing sit behind the ambient view.
The right abstraction — agent trajectories as state-transition trees, timelines and failure chains. Diagnostic and provider-agnostic rather than cinematic.
Hierarchical trace tree + failure-onset localization
A research tracing architecture that parses heterogeneous run artifacts, reconstructs the full state transition history as a hierarchical trace tree with persistent memory, and localizes where a failure began and how it propagated downstream. Ships with the CodeTraceBench benchmark.
React components for visualizing agent traces
An open-source React component library (by Evil Martians) that turns raw OpenTelemetry traces into interactive timeline and tree views — LLM calls, tool executions and subagents as a hierarchical timeline that instantly reveals loops, errors and cost spikes.
Open-source multi-step agent trajectory analysis
An open-source LLM observability platform whose multi-step trajectory analysis helps you understand complex agent behavior. Built on OpenInference/OpenTelemetry with framework-native tracing across 40+ integrations; self-hostable on a single node.
Open-source LLM & agent tracing
An open-source observability platform that captures the full context of each request — prompts, responses, tool calls and their relationships — and renders agent runs as trees with parent runs, tool calls, subagents and retries.
Open-source tracing & agent-execution graphs
Comet's open-source LLM-observability platform: it captures traces across 60+ frameworks (LangGraph, LangChain, LlamaIndex, CrewAI, Google ADK, OpenAI, Anthropic) and renders agent runs as execution graphs and span trees alongside evaluation dashboards.
Session replay & event graphs for agent frameworks
An observability SDK + dashboard with session replay, time-travel debugging and multi-agent event graphs across the broadest framework set — CrewAI, AutoGen/AG2, OpenAI Agents SDK, LangChain, LlamaIndex and more. The event graph is about as "Gource-like" as the observability tools get.
Agent traces + synced browser session replay
An open-source, OpenTelemetry-native observability platform built for agents: a conversation-style trace view of the run plus browser session replay synced to the timeline, so you can watch what the agent saw and did in step. Works with the Vercel AI SDK, Browser Use, Stagehand, LangChain and more.