
agent-workspace
Agent Workspace: run Claude Code and Codex agents in parallel from one meeting-style desktop app, with always-on voice routing.
Install with your AI
Paste into Claude Code, Cursor, or any agent — it reads the repo and wires the tool into your project.
Install and set up agent-workspace (git-clone project) into my current project. Found on https://claudeers.com/agent-workspace Repo: https://github.com/Flitternie/agent-workspace Homepage/docs: — Detected install method: git-clone → git clone https://github.com/Flitternie/agent-workspace Category: devops. Platforms: cli, desktop, web. Read the repo's README for exact setup and env vars, then install it and wire it into my project. Claudeers Health Verdict: unknown; community-verified: false. Confirm the source before running anything.
git clone https://github.com/Flitternie/agent-workspace
// compatibility
| Platforms | cli, desktop, web |
|---|---|
| Operating systems | — |
| AI compatibility | claude |
| License | — |
| Pricing | open-source |
| Language | TypeScript |
Agent Workspace
A desktop workspace for running several coding agents (Claude Code and Codex) in parallel, designed like a meeting rather than a set of terminals.
- Overview as a meeting. Every agent is a card that reports what it did and what it needs from you. Questions and permission requests stay on their card until you answer, however long you spend with another agent.
- A main agent to talk to. Unless you pick an agent, you talk to a main agent (your own Claude Code or Codex) that sees every agent, passes instructions on, starts new agents or branches one off with its context (a fork, to try approaches side by side), answers status questions and collects what they need from you, so you can run the whole team from one conversation. It never edits code itself and cannot approve another agent's permission request on its own.
- Talk to them. The microphone stays open like a call. Say "Alice, add a regression test" or "小张,迁移先停一下" and it goes to that agent without changing your view. Short pauses join one message; a leading "stop / 等一下" halts the agent at once.
- Steer or interrupt. A message to a busy agent joins its current turn by default. ⌘Enter, or a spoken "stop", interrupts it and starts over.
- Uses what you already have. Agents run through the Claude Code and Codex CLIs installed on your Mac, with your existing logins. Permissions default to each tool's own auto mode (never bypassed), and can be set per agent.
- Speech, cloud or local. OpenAI transcription, or on-device models (SenseVoice, Whisper, Paraformer) that are downloaded on demand and never send audio anywhere.
- Also: per-task git worktrees, continuing existing Claude Code / Codex sessions from any folder, unified diff / terminal / plan rendering for both providers, Markdown, scoped memory, and English / 中文 UI.
Run
Requirements: macOS (Apple Silicon tested), Node 22, and the claude and/or codex CLIs, signed in. An OpenAI key (OPENAI_API_KEY, or an openai.key file next to the app) is optional; without it, use a local speech model.
npm run setup # installs into app/node_modules; npm and Electron caches stay in app/.cache
npm start # the real workspace (data in app/.data)
npm run demo # a separate demo with three mock agents (data in app/.data-demo)
Everything the app writes stays inside app/ (data, worktrees, downloaded models, caches). It does not change ~/.claude, ~/.codex or any global configuration.
Install
Download the latest installer from Releases:
- macOS (Apple Silicon): open the
.dmgand drag Agent Workspace to Applications. The app is not notarized by Apple yet, so the first time, right-click it in Applications and choose Open (or runxattr -dr com.apple.quarantine "/Applications/Agent Workspace.app"). - Windows (x64, experimental): run the
.exeinstaller. SmartScreen may warn about an unknown publisher: choose More info → Run anyway.
You still need the claude and/or codex CLI installed and signed in; the app uses them, not its own copies.
Release
Bump version in app/package.json on main (cd app && npm version minor --no-git-tag-version, commit, push). GitHub Actions tags the commit v<version>, runs the unit tests, builds the DMG and the Windows installer, and publishes the release with generated notes once both are attached. Pushes that keep the version release nothing. To build locally: npm --prefix app run dist:mac (output in app/release/; the .dmg file also gets the app icon in Finder).
Repository
The app (Electron + React + TypeScript) lives in app/. See app/README.md for scripts, tests, layout and third-party credits.
// faq
What is agent-workspace?
Agent Workspace: run Claude Code and Codex agents in parallel from one meeting-style desktop app, with always-on voice routing.. It is open-source on GitHub.
Is agent-workspace free to use?
agent-workspace is open-source, so it is free to use.
What category does agent-workspace belong to?
agent-workspace is listed under devops in the Claudeers registry of Claude-compatible tools.
// embed badge
[](https://claudeers.com/agent-workspace)
// retro hit counter
[](https://claudeers.com/agent-workspace)
// reviews
// guestbook
// related in DevOps & CI/CD
⭐AI-driven public opinion & trend monitor with multi-platform aggregation, RSS, and smart alerts.🎯 告别信息过载,你的 AI 舆情监控助手与热点筛选工具!聚合多平台热点 + RSS 订阅,支持关键词精准筛选。AI…
Use Claude Code as the foundation for coding infrastructure, allowing you to decide how to interact with the model while enjoying updates from Anthropic.
Professional Antigravity Account Manager & Switcher. One-click seamless account switching for Antigravity Tools. Built with Tauri v2 + React (Rust).专业的 Antig…
Roadmap to becoming an ASP.NET Core developer in 2026