Chat Hub — AI skill for Claude Code
Desktop cockpit for AI coding agents: Claude Code / Codex / Grok / OpenCode side by side — worktree-per-session, surfaces dock (files/terminal/diff/browser), MCP browser control, permission broker.
How to install Chat Hub
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open KolyshkinYafim/chat-hub and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Chat Hub does
Desktop cockpit for AI coding agents: Claude Code / Codex / Grok / OpenCode side by side — worktree-per-session, surfaces dock (files/terminal/diff/browser), MCP browser control, permission broker. Electron + React + TS
Alternatives in AI
- Adopt AI Properly — Workflow recipe — the org-side AI adoption arc, policy to proof, by chaining 4 skills 1.3k ★
- Browserwing — BrowserWing turns your browser actions into MCP commands Or Claude Skill, allowing AI agents to control browse 1.2k ★
- Dinotty — Multi-device terminal server for AI coding agents 643 ★
README
Chat Hub
[](https://github.com/KolyshkinYafim/chat-hub/actions/workflows/ci.yml)
Chat Hub is a desktop cockpit for AI coding agents: one window that runs Claude Code, Codex, Grok and OpenCode sessions side by side, each in its own project and, optionally, its own git worktree. Around the transcript it adds the working context a coding session actually needs — files, terminal, browser, diffs, a project board — and it stays wired into Session Monitor, the Swift menu-bar app in the sibling repo, so session status and permission prompts follow you outside the window.
Electron · TypeScript · React · pnpm — macOS first.

Features
- Multi-provider sessions — adapters for Claude Code, Codex, Grok and OpenCode CLIs (auto-detected on PATH), plus a mock provider for UI work. Streaming transcripts with tool cards, diffs, plans, mermaid diagrams and usage stats; honest status driven by process events, never a stuck "Working".
- Surfaces dock — per-session panels next to the chat: file browser with code editor and image/media/PDF preview, real terminal (node-pty + xterm), git diff / source control with per-hunk staging, project board (todos and notes the agent can edit too), embedded browser, and session history.
- Agent browser control — any of the four CLIs can drive the Browser surface over MCP: open pages, read them as ref-tagged trees, click, type, screenshot, read console and network logs — in the same webview the user is watching. See docs/browser-control.md.
- MCP manager — Settings → Connections edits a per-project server list (
.chathub/mcp.json) and materializes it into each provider's native config format. - Voice dictation — hands-free composer input through Handy, a local dictation app; Chat Hub never touches audio.
Related Skills
Mission Control
Local web cockpit orchestrating three AI coding engines — Claude Code (tech lead), GLM-5.3-Flash via z.ai (jun
Vibe Explain
Cognitive debt map for AI-generated code. Surfaces opaque blocks you don't fully own or understand, generates
Agent Eval Hub
Multi-provider + cross-device reliability harness for AI agents. Runs the same suite across Claude, OpenAI, Ge
Dux
Dux is a terminal UI that lets you run multiple AI coding agents side by side, each in its own git worktree, w
AI Usage Dashboard
Local-first quota cockpit for Claude Code, Codex, Grok Build, and DeepSeek API balance on Windows, macOS, and
Registry Broker Skills
AI agent skills for the Universal Registry - search, chat, and register 72,000+ agents across 14+ protocols. W
Related Agents
Protocol Parity
Cross-references the lockstep surfaces a protocol must touch (TUI ScreenId, admin route, live broker/session w
Agent Of Empires
Claude Code, OpenCode, Mistral Vibe, Codex CLI, Gemini CLI Coding Agent Terminal Session manager via tmux and
Omg Tracer
Read-only runtime/path tracer for oh-my-grok. Follow control flow and cite file:line.