Track token cost and spend across every AI coding agent session on a machine.
Review a syntax-highlighted diff of everything an AI agent changed before committing.
Watch live error timelines and tool-call latency for AI coding agents.
Manage git and Docker for a project from a single cockpit alongside agent telemetry.
| sirallap/agentglass | yacuo/check-cc | katipally/openlive | |
|---|---|---|---|
| Stars | 192 | 190 | 189 |
| Language | TypeScript | TypeScript | TypeScript |
| Setup difficulty | moderate | easy | easy |
| Complexity | 3/5 | 2/5 | 3/5 |
| Audience | developer | general | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires Bun to run the server and wiring up Claude Code hooks or an OpenTelemetry exporter for other providers.
agentglass is a real time dashboard and workspace for watching AI coding agents work across every project on your machine. It connects to Claude Code through hooks, or to other providers such as OpenAI Codex, Gemini CLI, Bedrock, LangChain, or LiteLLM through OpenTelemetry, and shows every agent session, tool call, token used, and dollar spent as it happens. It tracks cost, how long tools take to respond, error timelines, and full session histories, and this data is saved so it survives a page reload rather than only existing as a live stream. It watches every Claude Code session on the machine automatically, without needing to be launched separately for each project, and it also scans past transcript files so history from every project shows up as soon as the dashboard opens. You can narrow its focus to a single project and its related git worktrees, to a whole folder of projects, or leave it covering the entire machine, and that choice is remembered between launches. Beyond just watching, agentglass includes a full workspace built from several panels styled like popular terminal tools. There is a syntax highlighted diff viewer for reviewing every file change an agent makes, with a one click AI generated explanation of the whole change set and the ability to turn a review directly into a commit. A source control panel modeled on lazygit lets you stage, unstage, and commit changes across any repository it discovers. There is also a Docker panel modeled on lazydocker, a real terminal running an actual shell process rather than a simulation, and a chat panel for driving local Claude Code sessions. All of this sits behind a single keystroke, with a persistent notification strip at the top showing pending git pushes and pulls, live shells, and sessions waiting for your attention. The project offers 22 visual themes and can run either in a browser or as a native desktop app built with Electron. The backend runs on Bun with SQLite, and the frontend uses React, Vite, and Shiki for code highlighting. It is released under the MIT license. The full README is longer than what was shown.
A real-time dashboard and workspace for monitoring AI coding agents like Claude Code across every project.
Mainly TypeScript. The stack also includes TypeScript, Bun, SQLite.
MIT licensed, free to use, modify, and distribute for any purpose including commercial use.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.