M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
C

Claude Code Haha (cc-haha)

Community
Not a standard server
Local-first desktop workspace for Claude Code and AI coding agents — an MCP client with a built-in visual manager, not an MCP server itself.

This is not a typical connect-and-use MCP server; read the source repository before using it.

How to install ↓ Compare GitHub source repository ↗
Category
Dev Tools #258 of 438
Stars
★ 15.3k Very popular
Runtime
Prebuilt binary · Bun
Credentials
API key / credential required
License
Other / unspecified
Last commit
50FMRS · D

cc-haha is a mature open-source Claude Code desktop workspace covering multi-session management, Git Worktrees, diff review, multi-model support, Computer Use, and IM integration. Based on the available README and materials, however, it is not itself an MCP server that other MCP clients can connect to — it's an MCP client application with a built-in visual manager for connecting to third-party MCP servers.

Strongest · Maintenance 15/20 Weakest · Reliability 4/20

Reliability
4/20
Security and permissions
8/20
Maintenance
15/20
Documentation
14/20
Setup experience
9/20
Why each score
Reliability 4/20
No server manifest was cached, and the README lists no MCP tool names, parameters, input/output schemas, or error-handling conventions. The repository reads more like a desktop client that can 'add/manage' external MCP servers (the README describes a 'Visual MCP manager' for adding third-party MCP servers) rather than clearly exposing itself as an MCP server with tools — this fundamental ambiguity cannot be resolved from the static material provided. No visible CI workflows or committed tests demonstrate the init + tool-list handshake succeeding. Per the static calibration rule, reliability cannot exceed 12 without verifiable execution evidence, and given the severe evidence gap it is scored even lower at 4.
Security and permissions 8/20
Several permission/data-boundary concerns are present: 'Computer Use' lets the agent take screenshots, click, type, and control desktop apps (stated as gated 'after authorization' but with no detail on authorization granularity or revocation); one of the 'five permission modes' explicitly allows 'skip permissions,' bypassing confirmation for risky operations when selected; IM integrations (WeChat, Feishu, DingTalk, Telegram, WhatsApp) and H5 remote access allow driving the agent remotely through third-party messaging channels, expanding the attack surface and potential for credential/session hijacking. The README does not describe how API keys/credentials are stored, least-privilege design details, or data-egress boundaries, nor any security-response channel. Because risks are visible but explicit default safeguards, confirmation mechanics, and data-flow disclosure are incomplete (though authorization/confirmation is nominally mentioned), this falls short of a red-line 0-4 score but is scored in the low-middle range.
Maintenance 15/20
The repository is not archived, uses the MIT license, has a high star count (13,926), and 135 open issues indicate ongoing user engagement; the README lists sponsors, a contact email, a dedicated documentation site (cchaha.ai), and a Feishu user community, showing clear ownership and communication channels. However, no release history, changelog, or dependency-update cadence is provided, and there is no described security-response process, so recent commit/release activity cannot be verified directly — preventing a top score.
Documentation 14/20
Documentation is fairly extensive, with layered links covering getting started, desktop features, IM integrations, CLI usage, and internal architecture. However, because the server manifest was not cached, no concrete MCP tool parameters, invocation examples, rate limits, or cost disclosures appear anywhere — core content expected for an MCP server review is missing. A troubleshooting doc is referenced but its actual content is not present in the supplied material, so its quality cannot be verified, resulting in a deduction.
Setup experience 9/20
Two install paths are given: downloading a desktop installer, or running from source via `bun install` + `.env` configuration + `./bin/claude-haha`, with clearly described steps and known caveats for macOS signing/notarization and Windows SmartScreen. However, there is no configuration example for connecting this project as an MCP server to a mainstream client (e.g., a mcp.json snippet), and no CI/test evidence that these steps reproducibly work. Per the static calibration rule, setup is capped at 15 without such evidence, and it is scored in the low-middle range given the missing client-integration example and verifiable execution proof.

Static review · not runListed 2026-08-06

Read the FMRS scoring method →

Fit and risk

What it can accessReads local filesWrites / deletes local filesRuns commands or codeUses the networkControls a browser

Best for

  • Developers already using Claude Code or similar coding agents who want a graphical desktop workspace
  • Users who need to manage multiple model providers (Claude/ChatGPT/Grok/local models) and multiple MCP server connections in one place
  • Teams or individuals who want to monitor and approve agent actions remotely through chat apps

Not for

  • Anyone looking for an MCP server that Claude Desktop, Claude Code, or another MCP client can connect to and call specific tools on
  • Headless server deployments where a desktop Electron GUI isn't wanted or usable
  • Users unwilling to install a local Electron desktop app or wary of installing skills from a third-party marketplace

Required permissions

  • Local file system access for Git Worktree operations, diff review, and project files
  • Network access to reach model provider APIs and IM channels (WeChat, Feishu, DingTalk, Telegram, WhatsApp)
  • Computer Use permission (once authorized) allowing the agent to take screenshots, click, type, and control desktop apps
  • A 'skip permissions' mode is available among the five permission levels, letting the agent run commands and tool calls without per-action approval

Risks and side effects

  • Computer Use can directly control the desktop once authorized; an overly permissive permission mode raises the chance of unintended actions
  • The app stores API keys for multiple model providers locally, concentrating sensitive credentials on one device
  • The skill marketplace allows installing third-party skills from ClawHub/SkillHub; source and safety status are shown, but supply-chain risk remains
  • Unsigned draft or temporary builds require manual one-time approval to run, which can be risky if the build source isn't trusted
  • As an MCP client, it can be pointed at arbitrary third-party MCP servers, so users must independently vet the trustworthiness of any server they add

Setup

Before you start

Runtime：Prebuilt binary · Bun

Download the macOS/Windows/Linux desktop installer from GitHub Releases and install it; on first launch, configure your model provider, API key, and default model in Settings. Public macOS releases are signed and notarized, but draft or unsigned temporary builds may still require one-time manual approval; unsigned Windows installers may trigger SmartScreen — click 'More info' then 'Run anyway'. To run the underlying CLI/server from source for debugging: bun install; cp .env.example .env; ./bin/claude-haha. Because this project is not itself a connectable MCP server, there is no standard MCP client configuration (e.g., an entry in claude_desktop_config.json) for it.

Check that it works

Install the desktop app from Releases, launch it, and configure a model provider and API key in Settings; creating a session and receiving a model reply confirms the app is working.

Troubleshooting

  1. On macOS, a one-time manual approval prompt usually indicates a draft or unsigned temporary build — see the desktop install guide
  2. On Windows, if SmartScreen blocks the installer, click 'More info' then 'Run anyway'
  3. When running from source, confirm bun install completed and .env is configured correctly (see the environment variables doc)
  4. If model calls fail, check the model provider, API key, and default model settings in the app's Settings panel
  5. For a specific MCP server connection issue, check that server's scope (project/shared/global) and transport configuration in the visual MCP manager

Things to try

Once connected, you can ask your AI assistant things like:

  • Open this repo's dev branch in an isolated Worktree and start a new session
  • Review the diff of this turn's changes and undo the whole turn
  • Preview the page you just edited in the built-in browser
  • Move this session to my phone via the H5 QR code

Use cases

Managing multiple Claude Code / coding-agent sessions on the desktop with tabs, project switching, and session history
Launching tasks against a Git branch or an isolated Worktree, then reviewing changes file by file as syntax-highlighted diffs
Adding and editing third-party MCP servers (STDIO/SSE/Streamable HTTP) through the built-in visual MCP manager
Continuing or approving running tasks remotely via Feishu, Telegram, WeChat, DingTalk, WhatsApp, or a phone H5 page
Letting an authorized agent take screenshots, click, and type to control desktop apps via Computer Use

Overview

Claude Code Haha (cc-haha) is a local-first, cross-platform desktop app for macOS, Windows, and Linux built around Claude Code and other coding agents. It provides multi-session workspaces, Git branch/Worktree launch, file-by-file diff review, a built-in browser preview, GUI permission approval, support for multiple models (Claude, ChatGPT, Grok, DeepSeek, Kimi, Zhipu GLM, and local endpoints), image generation, task-aware desktop pets, and remote access via WeChat, Feishu, DingTalk, Telegram, WhatsApp, and H5. Although the repository is tagged 'mcp', the README describes the app's MCP-related capability as a 'visual MCP manager' for adding and editing third-party MCP servers (STDIO / Streamable HTTP / SSE, with project, shared, or global scope) — there is no evidence in the source material that cc-haha itself exposes any MCP tools or acts as a server that other MCP clients can connect to. In other words, cc-haha functions as an MCP client/management surface, not an MCP server.

Similar servers

Context7 80 · B

Upstash's official server providing up-to-date third-party library docs for AI coding assistants

★ 64.2k · Tools 2 Compare with this →
Compare with other MCP servers

Source revision 8be7d18acba9 Data synced 2026-11-05 Read the FMRS scoring method