Claude Code Haha
A cross-platform desktop workspace for Claude Code and multi-model coding tasks.
- Source repo
- NanmiCoder/cc-haha
- Stars
- ★ 15k
- Last updated
- 1d ago
- License
- NOASSERTION
- Primary language
- TypeScript
- FA score
- 44/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platformChatGPT · Claude Code · Claude API
- Cost
- Free software; you pay for model usage
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- A Claude Code user managing several repositories and sessions who wants to inspect task progress and tool activity in one desktop workspace.
- Not a fit if
- Developers who want a terminal-only workflow and no Electron desktop app
- Users who refuse pay-per-use model API keys and want only free usage
- Source review
- 44/100 · Major gaps
What does this agent do, and when should you use it?
Claude Code Haha is a desktop Claude Code workspace for macOS, Windows, and Linux, with a source-run CLI path. Its Electron and React interface brings sessions, project switching, branch and Git Worktree launch, file changes, and syntax-highlighted diff review into one workspace. Users configure a model provider, API key, and default model, with documented sign-in or configuration paths for Claude, ChatGPT, Grok, several presets, LM Studio, and Ollama. The app also documents Computer Use, phone-browser H5 access, IM relays, scheduled tasks, usage statistics, a skill marketplace, and task-aware desktop pets. It fits developers who want a graphical control surface for local coding-agent sessions and isolated worktrees rather than a terminal-only workflow.
In the desktop app, a user creates or switches projects and sessions, selects a repository branch, then chooses the current working tree or an isolated Worktree. The workspace lists changes from the current turn; users can open files for syntax-highlighted diffs or undo the entire turn. Five permission modes let users approve risky commands, tool calls, and follow-up questions in the GUI, while Computer Use can take screenshots, click, type, and control desktop apps after authorization. The skill marketplace discovers, previews, and installs third-party skills from ClawHub / SkillHub with source and safety status, and scheduled tasks run in their own sessions. The documented source CLI flow runs bun install, cp .env.example .env, and ./bin/claude-haha.
- A Claude Code user managing several repositories and sessions who wants to inspect task progress and tool activity in one desktop workspace.
- An engineer working across branches who wants to choose between the current tree and an isolated Git Worktree before launching a task.
- A technical lead reviewing agent edits who needs file-by-file syntax-highlighted diffs and the option to undo an entire turn.
- A developer following a long-running task away from the desk who wants to continue a session or approve actions through H5, Telegram, Feishu, WeChat, DingTalk, or WhatsApp.
- A team member configuring coding models across hosted and local options, including Claude, ChatGPT, Grok, LM Studio, and Ollama.
How do you install or deploy this agent?
Desktop app: download the macOS, Windows, or Linux installer from GitHub Releases. On first launch, configure the model provider, API key, and default model in Settings. Public macOS releases require signing and notarization; draft or unsigned temporary builds may need one-time manual approval. Unsigned Windows installers may trigger SmartScreen; select “More info” and then “Run anyway.” Source CLI: install Bun, then run bun install, cp .env.example .env, and ./bin/claude-haha; configure provider credentials in .env.
How do you use this agent?
In the desktop app, first set the model provider, API key, and default model, then create or open a project and session. When starting work, choose a branch and either the current working tree or an isolated Worktree; review resulting workspace diffs file by file and undo the turn when necessary. For remote follow-up, scan the QR code to continue in a phone browser or pair a supported IM channel for conversation, project switching, and approvals. For the source CLI path, configure .env and run ./bin/claude-haha.
What are this agent's strengths and limitations?
- Combines session management, project navigation, branch/Worktree selection, and file-level diff review in a cross-platform desktop workspace.
- Offers five permission modes and GUI approval for risky commands, tool calls, and follow-up questions.
- Documents paths for Claude, ChatGPT, Grok, multiple provider presets, and local-model options through LM Studio and Ollama.
- H5 remote access plus Telegram, Feishu, WeChat, DingTalk, and WhatsApp relays support mobile task follow-up and approvals.
- Computer Use, scheduled tasks, the skill marketplace, and the session activity panel extend the workflow beyond a standard terminal session.
- The desktop app uses Electron; the source CLI explicitly requires Bun, and Worktree usage depends on a Git workflow.
- Initial use requires a model provider, API key, and default model; provider pricing and quotas are not documented.
- Computer Use can control desktop applications; despite authorization requirements, users must choose permission modes carefully.
- Unsigned temporary macOS builds and Windows installers can require extra operating-system security approval.
- The marketplace installs third-party skills; source and safety status are shown, but the materials do not describe a separate audit guarantee for every skill.
How does this agent compare with similar options?
The materials name cc-switch as a reference for model-provider configuration. Claude Code Haha's documented scope additionally includes desktop sessions, Worktrees, diff review, remote access, and Computer Use.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Claude Code Haha This agent | 44 · Major gaps | Desktop appFree + model costs | ★ 15k | 1d ago | TypeScript | ChatGPT · Claude Code · Claude API |
| Poracode | 55 · Major gaps | Desktop appFree + model costs | ★ 121 | today | TypeScript | ChatGPT · Claude Code |
| Superset | 52 · Major gaps | Desktop appFreemium | ★ 16k | today | TypeScript | Codex · Claude Code |
| Ateam Agent Workbench | 70 · Some gaps | Desktop appFree + model costs | ★ 16 | today | TypeScript | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows permission modes, GUI approval, undo, code signing policy, and privacy docs, but lacks implementation details or security audit. Deductions: least privilege, user confirmation, data flow transparency, sensitive data handling, dependency security, external effects, rollback, and source attribution are only partially supported, lacking in-depth evidence.
Extensive test scripts and test files indicate attention to consistency, but no test results or coverage data provided. Many dependencies, but no availability guarantees. Failure messages: troubleshooting docs exist, but no specific error handling examples. Deductions: dependency availability and failure messages insufficiently evidenced.
Multi-platform, multi-model, multi-IM integration, broad scenarios. Capability boundaries documented (permission modes, feature list), but no explicit limitations. Trigger precision: command reference and env docs, but no specific trigger conditions. Environment fit good, multi-platform support. Deductions: capability boundaries and trigger precision insufficiently evidenced.
Information architecture clear, docs directory and feature list. Install notes detailed, including download, configuration, and signing. Naming stable, but version number anomalous (999.0.0-local). Examples and FAQ present, but known limitations not explicitly listed. License MIT, but copyright year 2026 may be inaccurate. Versioning changelog not provided. Maintenance responsibility by author, but not explicit. Deductions: naming stability, known limitations, versioning changelog, and maintenance responsibility insufficiently evidenced.
Output usability high, rich features and docs. Marginal value high, multi-model, multi-platform support. Cost-benefit: sponsors and donation options, but no pricing or cost analysis. Deductions: cost-benefit insufficiently evidenced.
Claims supported by docs, but no traceable links or evidence. Cross-source corroboration limited, only internal docs. Fact-inference separation unclear, some claims may be inferred. Deductions: claim traceability, cross-source corroboration, and fact-inference separation insufficiently evidenced.
- Anomalous version number (999.0.0-local) may indicate unreleased state.
- Copyright year 2026 may be inaccurate.
- Many dependencies, but no security audit or vulnerability scanning evidence.
- Security features like permission modes and data flow transparency lack implementation details.
- No explicit list of known limitations.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
v0.7.7 Claude Code Haha v0.7.7
Release notes excerpt
Claude Code Haha v0.7.7 本次让手机 App 在后台也能收到提醒，不再需要自建推送服务器，手机和平板的浏览器也能直接打开手机 App 同一套界面；桌面端新增输入建议、任务完成后的系统操作、会话归档、MCP 共享到多个项目，语音输入可选本地模型并支持 AI 纠错和个人词库，Windows 和 Linux 的 Computer Use 改用原生运行组件。同时修复 Agent Teams 成员上下文溢出卡死、中转站上下文超限无法恢复、升级安装和本地索引重建时的启动失败等一批问题。安装包改用更直观的文件名，Windows 安装包默认未签名。 核心功能 - 手机 App 后台提醒（#1780）：离开 App 后的审批、完成和出错提醒不再依赖发布者的推送密钥，也不需要自建服务器。iPhone 由 Expo 推送服务转发系统通知，电脑需要能访问 exp.host；Android 由通知栏里常驻的「正在值守」向电脑拉取，不依赖 Google 服务，部分国产手机需要在 App 的「设置 → 通知」里允许后台运行。手机设置里会显示后台连接和电池优化状态。 …View on GitHub ↗ -
v0.7.6 Claude Code Haha v0.7.6
Release notes excerpt
Claude Code Haha v0.7.6 本次带来 cc-haha 手机 App（Android / iPhone），手机连接也改用 Cloudflare：离开电脑也能在手机上看进度、审批、派任务。桌面端新增运行任务时防止系统休眠、在对话里管理定时任务、编辑并重发被停止的提示，供应商方面加入 OpenRouter 预设、按供应商设置代理和官方账号刷新模型列表；同时修复 Agent Teams 卡在收尾、长会话滚动位置、大图请求等一批问题。 核心功能 - cc-haha 手机 App（#1688，#1698，#1710，#1715，#1724，#1725，#1727，#1729，#1738，#1741，#1743）：在桌面「手机连接」里扫码配对自己的电脑后，可以在手机上看流式回复、审批工具权限和回答提问、发新任务、创建定时任务、浏览文件和改动。语音输入由电脑上已配置的语音服务转写。Agent 团队的方案可以在手机上审阅，逐个成员调整模型后启动；启动后按成员查看进度和消息，也能从手机停止团队。经 Cloudflare 连接时，工具截图按引用传输，弱网下重连更快。任务始终在电脑上执行，手机只和你的电脑通信。App 界面目前只有中文。 …View on GitHub ↗ -
v0.7.5 Claude Code Haha v0.7.5
Release notes excerpt
Claude Code Haha v0.7.5 本次带来 跨会话模型分工、重做后的 Token 用量页 和 实验性的精简 Agent 模式，工作区也支持给 Word、PDF、表格加批注；同时修复 OpenCode Go 的 Gemini 模型不可用、macOS 全屏、缓存命中率统计、Windows 数据迁移权限预检等一批问题。 核心功能 - 跨会话模型分工（#1626，#1628，#1629，#1631）：一个会话可以把任务分给若干子会话，每个子会话单独指定供应商和模型；也可以先看一份方案，再决定要不要启动。分工列表移到了「活动」栏，展开就能看到每个会话用哪个模型、跑到哪一步。已启动的会话也能换模型，可以选「本轮结束后切换」或「立即暂停并切换」，切换没生效前会话照常用原来的模型把当前轮跑完；一次最多修改 8 个已启动会话。 …View on GitHub ↗