CCEM

The Control Center for Your AI Coding Assistants

English | 中文

Table of Contents

• Why ccem
• What ccem Does
• Competitive Landscape
• Quick Start
• CLI
• Desktop App
• Coming Soon
• Data Storage
• Tech Stack
• Contributing
• License

Why ccem

After using Claude Code for a while, some things just get annoying.

You want KIMI for frontend, Opus for architecture, DeepSeek for scripts — but every time you switch models you're manually export-ing env vars. Eight terminal tabs later, no idea which is connected to what.

Permissions too. Approving every command gets old. But --dangerously-skip-permissions feels reckless — what if it deletes your .env?

Then there's the money question. How much have you spent this month? Claude Code won't tell you. Codex won't either. The combined bill across both? Nobody knows.

And the real kicker — you're out getting coffee, need Claude to fix something, but your computer's at home. You're stuck.

Or maybe you want Claude to run tests every night, review PRs, push results to your phone. Not a crazy ask, right?

So I built ccem. Environment switching, permission modes, multi-model sessions, usage analytics, cron automation, remote control from your phone — one tool, all covered.

What ccem Does

Two flavors. Pick whichever fits.

Both share the same config file (~/.ccem/config.json). Change an environment in Desktop, CLI picks it up instantly — and vice versa.

Competitive Landscape

Quick Start

CLI (Terminal)

npx ccem              # Interactive menu — no install needed

Or install globally:

npm install -g ccem
ccem add kimi         # Add an environment (auto-fills URL and models)
ccem use kimi         # Switch to it
ccem dev              # Launch Claude Code in dev permission mode

Desktop

Download from GitHub Releases:

• macOS: .dmg (Apple Silicon / Intel)
• Windows: .exe installer (x64)

Launch, add an environment, pick Claude or Codex, hit go.

CLI

Environment switching, permissions, usage stats, and skill installation — all from your terminal.

Install

npm install -g ccem
# or: pnpm add -g ccem
# or just: npx ccem

Environment Management

ccem              # Interactive menu
ccem add kimi     # Add with preset (auto-fills URL + models)
ccem use kimi     # Switch environment
ccem ls           # List all environments
ccem current      # Show active environment
ccem env          # Output export commands (pipe-friendly)
ccem env --json   # JSON format
ccem run <cmd>    # Run command with env vars injected
ccem del <name>   # Delete environment
ccem rename <a> <b>
ccem cp <src> <dst>

Built-in Presets

Official environment defaults to claude-sonnet-4-5-20250929 + claude-haiku-4-5-20251001.

Shell Integration

After ccem use, env vars won't update in your current shell. Add this to ~/.zshrc:

ccem() {
  command ccem "$@"
  local exit_code=$?
  if [[ $exit_code -eq 0 ]]; then
    if [[ "$1" == "use" || -z "$1" ]]; then
      eval "$(command ccem env)"
    fi
  fi
  return $exit_code
}

Then source ~/.zshrc.

Permission Modes

Six presets between "approve everything" and "approve nothing."

ccem yolo / dev / readonly / safe / ci / audit   # Temporary (reverts on exit)
ccem setup perms --dev                            # Permanent
ccem setup default-mode --dev                     # Set default
ccem --mode                                       # Show current

Usage Analytics

ccem usage          # Interactive with calendar heatmap
ccem usage --json   # Machine-readable

Parses JSONL logs from ~/.claude/projects/ to calculate token usage and costs. Price data fetched from LiteLLM and cached locally.

Skill Management

ccem skill add              # Interactive picker (Tab to switch groups)
ccem skill add <name>       # Install preset
ccem skill add <github-url> # Install from GitHub
ccem skill ls               # List installed
ccem skill rm <name>        # Remove

Official presets: frontend-design, skill-creator, web-artifacts-builder, canvas-design, algorithmic-art, theme-factory, mcp-builder, webapp-testing, pdf/docx/pptx/xlsx, brand-guidelines, doc-coauthoring

Curated: superpowers, ui-ux-pro-max, Humanizer-zh

Remote Config

Share API configurations across your team with encrypted transport.

ccem load https://your-server.com/api/env --key YOUR_KEY --secret YOUR_SECRET

Server code lives in server/. AES-256-GCM encryption with authenticated envelope (v2), rate limiting, and hot-reload. See server/ in the repo for full deployment instructions.

CLI Command Reference

Desktop App

A native desktop app built with Tauri 2.0 — real Rust + React, not an Electron wrapper. Runs on macOS and Windows.

On top of everything the CLI offers, Desktop brings an integrated workspace, dual-engine support, remote control, cron automation, and a tray mini-dashboard.

Install

Download from GitHub Releases:

• Windows x64: CCEM Desktop_*_x64-setup.exe
• macOS (Apple Silicon / Intel): .dmg — macOS 10.15+

Workspace — Your Command Center

The Workspace is where sessions live. It's not just a launcher — it's a full control surface.

Prompt Composer — Launch sessions with a rich input: $skill tokens, /commands, @file references, image attachments, model/provider/effort selectors. Everything you need before hitting enter.

Slash Commands — ccem scans your installed Claude Code commands and .claude/commands/ directories, parsing YAML frontmatter. Available directly in the composer.

Global Search (Cmd+K) — Search across all past conversations and projects in real time. Find that thing Claude said three days ago.

Multi-Session Management — Run Claude Code in one window, Codex in another, DeepSeek in a third — simultaneously, each with its own environment and permission mode.

• Grid / list view toggle
• Per-session: project dir, environment, permission mode, PID, source (Desktop / CLI / Telegram / WeChat / WeCom / Cron)
• Per-session actions: focus, minimize, stop, close
• Orphan recovery: detect and take over unmanaged Claude processes

Session Review Drawer — Open any session to see a structured review: final assistant reply, changed files (git + SDK), extracted todos, tool evidence, generated artifacts (HTML/images/reports), subagent tracking.

Checkpoints & Rewind — Claude's file checkpoint system exposed in the UI. Create checkpoints, rewind to them, track rewind failures.

Subagent Tracking — ccem tracks subagent lifecycle with named scientific personas (Turing, Curie, Feynman, Hopper, etc.), each with unique symbols and accent colors.

Claude Code + Codex Dual Engine

Desktop supports both Claude Code and OpenAI Codex CLI as runtimes. The Dashboard launch panel has a dropdown — pick Claude or Codex, hit launch.

When Claude is selected, environment switching and permission modes work as usual. When Codex is selected, those controls hide — Codex manages its own config.

Both session types appear in a unified view, with proper icons and status indicators. Proxy Debug captures traffic from both engines.

ccem Desktop isn't just a Claude Code manager — it's a control center for your local AI coding assistants.

Remote Control — Telegram, WeChat & WeCom

Control Claude Code sessions on your computer from your phone. Unlike the official Claude mobile app (Anthropic subscription required), ccem works with any API key you've configured.

Telegram

Uses Forum Topics — each Topic maps to a project directory on your machine. One Topic, one project, one persistent session. Clean separation, no cross-talk.

1. Configure your Bot Token + allowed users in ChatApp
2. Bind each Forum Topic to a project directory, environment, and permission mode
3. Send messages from your phone — ccem spawns/reuses a local Claude session
4. Results stream back in real time, with optional tool call visibility

WeChat (微信)

Direct private chat mode — send a message, ccem spawns a headless Claude session, results stream back. QR code login, user allowlist, permission approval via /approve / /deny in chat.

WeCom (企业微信)

Full enterprise WeChat bot bridge with multi-bot management, WebSocket connectivity, admin/user permission separation, group chat support, @mention triggering, and cron-to-WeCom push notifications.

Feishu (Lark) integration — coming soon.

Cron Tasks — Scheduled Automation

Write a cron expression and a prompt — ccem runs Claude Code on schedule and pushes results to Telegram, WeChat, or WeCom.

• Templates: PR Review, Test Runner, Doc Generation, Security Audit, Changelog
• AI generation: Describe what you want in natural language, get a cron expression + prompt generated
• Auto-push: Results land in your bound chat app when done (or on failure)
• Run history: Status, duration, logs for every execution
• Next run preview: See when upcoming runs will fire
• Retry on failure: One-click re-run

Analytics — Claude + Codex, One View

GitHub-style usage statistics for both Claude Code and Codex, unified.

• Daily activity heatmap + token/cost trends by model
• Switch between Claude, Codex, or combined with one click
• Consecutive active days streak + trend arrows (up/down vs. last week)
• Share poster: generate your AI Coding weekly report

Tray Cockpit — System Tray Mini-Dashboard

A persistent mini-dashboard in your system tray, independent of the main window:

• Current environment and permission mode at a glance
• Today's token usage and cost
• 12-hour activity chart with interactive cursor tracking
• Provider breakdown: Claude / Codex / OpenCode
• Active sessions with status dots
• Upcoming cron tasks with expressions
• Quick-launch dock: Workspace, Sessions, Proxy Debug

Conversation History

Browse all past Claude Code and Codex conversations in one place. Filter by source (All / Claude / Codex), grouped by project directory. /compact segmentation boundaries supported.

Proxy Debug

Built-in API request debugging:

• Live traffic list: timestamp, method, URL, status code, size
• Request/response detail viewer with JSON formatting + SSE stream detection
• Separate upstream URL configuration for Claude and Codex

Environments & Skills

Same functionality as CLI, visual interface:

• Environments: Card-based list, one-click preset filling, remote config sync
• Skills: Streaming search as you type, one-click install/uninstall

App Auto-Update

Built-in Tauri updater — check, download, install updates from within the app. Progress tracking with status indicator.

External Control API

A local JSON-RPC server for programmatic control:

• ccem:// URI scheme deeplinks to sessions, events, and history
• Token-authenticated localhost endpoint
• create_session, list_sessions, send_input, open_session methods
• Enables bot-to-session binding and external tool integration

Settings

• Theme: light / dark / system
• Language: Chinese / English
• Default permission mode / working directory
• Terminal preference (iTerm2 / Terminal.app)
• AI-enhanced mode: use a selected environment for AI features
• Dependency check: auto-detect CLI / claude / codex / tmux

Keyboard Shortcuts

Coming Soon

• Feishu (Lark) — bot bridge integration
• Pet/Companion — pixel-art desktop companion that reacts to your session activity

Data Storage

Tech Stack

apps/cli/          CLI — Commander.js + Inquirer.js + Ink (React for CLI)
apps/desktop/      Desktop — Tauri 2.0 + React 18 + Rust
packages/core/     Shared logic — presets, types, encryption
server/            Remote config server

pnpm workspaces monorepo. Frontend: Vite, Tailwind CSS, Zustand, shadcn/ui, Recharts, GSAP. Backend: Rust + Tauri 2.0, window-vibrancy (macOS glassmorphism). i18n: Chinese / English.

Contributing

Issues and PRs welcome!

License

MIT