Instant containerized development with Claude Code by default, plus provider-aware agent tooling
PyPI • Documentation • GitHub
# Install Python package from PyPI
uv tool install cuti
# Launch Docker development environment with Claude
cuti containerThat's it! You now have a fully configured AI-powered coding environment with:
- ✅ Claude Code CLI pre-configured with Anthropic integration
- ✅ Persistent authentication for Claude API across sessions
- ✅ Python 3.11, Node.js 20, and essential dev tools
- ✅ Custom prompt showing
cuti:~/path $ - ✅ Auto-mounts current directory for seamless workflow
The Docker container provides isolated, reproducible AI-assisted development with Claude Code mode as the default. Each cuti container run refreshes Claude Code in the persistent container runtime when Claude Code mode is active, so subsequent containers pick up the newest available Claude CLI without a rebuild.
Build with multiple AI models and intelligent task management:
- Multi-agent orchestration - Claude API, Gemini integration
- Command queuing system with priority execution
- Read-only ops console - Launch with
cuti webto inspect provider readiness, queue state, recent activity, and workspace drift - Smart rate limiting - Automatic retry & backoff
- Task automation - Built-in todo system for AI agents
- Claude version switching - Easy CLI version management
- Agent providers - Run Claude Code mode by default, switch to OpenClaw mode with
cuti container --openclaw, and add Codex, Claude Code, OpenCode, or other providers throughcuti providers ... - Provider-aware setup - cuti mounts provider auth/config/skills, persistent CLI runtimes, and workspace instruction files for enabled tools
- Turnkey OpenClaw - Run the current OpenClaw gateway, channels, browser, plugins, and voice-call flows with
qt-openclaw ...orqt-OpenClaw ... - Claude chat history -
cuti historyshows transcripts and reopens old Claude Code sessions
Perfect for AI-powered development, automation workflows, and LLM orchestration.
Claude Code mode is the default for cuti container. OpenClaw is a separate mode, started explicitly with cuti container --openclaw or cuti container --claw.
cuti providers list
cuti providers doctor
# Default mode: Claude Code. This also refreshes Claude Code for future containers.
cuti container
# OpenClaw mode: OpenClaw only unless add-ons are explicitly enabled.
cuti container --openclaw
# Add-ons for OpenClaw mode are read from provider config.
cuti providers enable claude
cuti providers enable codex
cuti providers enable opencode
cuti providers enable hermes
cuti container --openclaw
# Provider setup/update helpers still run in the right mode automatically.
cuti providers auth claude --login
qt-openclaw onboard
qt-openclaw up
cuti providers update codex
cuti providers update openclaw
cuti providers update hermesWhen selected for the current mode, cuti handles the provider-specific container wiring for auth, config mounts, persistent CLI installation, and standard instruction files such as CLAUDE.md, AGENTS.md, .hermes.md, HERMES.md, SOUL.md, TOOLS.md, IDENTITY.md, USER.md, and HEARTBEAT.md. OpenClaw installs under ~/.cuti/provider-runtimes/openclaw, keeps runtime state under ~/.openclaw, and runs openclaw doctor --non-interactive during install/update/bootstrap. Hermes is currently marked experimental in cuti, but its state persists under ~/.hermes, including Hermes profiles under ~/.hermes/profiles. ~/.openclaw is mounted when available so hermes claw migrate --dry-run can inspect OpenClaw data before migration, and cuti providers update hermes follows Hermes' native hermes update path so bundled skills and profile state stay aligned with upstream behavior.
Use qt-openclaw or qt-OpenClaw for the current OpenClaw runtime.
qt-openclaw onboard
qt-openclaw up
qt-openclaw channels-login --channel whatsapp
qt-openclaw browser start
qt-openclaw plugins install @openclaw/voice-call
qt-openclaw voice-setup
qt-openclaw voicecall status
qt-openclaw dashboard --no-openThe command enables the OpenClaw provider automatically and runs inside OpenClaw mode, with ~/.openclaw, ~/.agents, and the persistent OpenClaw CLI runtime mounted for reuse.
OpenClaw's source-backed command families are exposed directly (models, mcp, sandbox, memory, wiki, approvals, nodes, devices, hooks, webhooks, tasks, cron, security, secrets, and more). For plugin commands added by future OpenClaw releases, use qt-openclaw run <command> ...; it forwards raw arguments to the installed OpenClaw CLI without requiring a cuti release.
| Guide | Description |
|---|---|
| Docker Container Setup | Container runtime, provider selection, mounts, and auth |
| Claude Authentication | Anthropic API & Claude CLI setup |
| Claude Account Switching | Manage multiple Claude accounts |
| Claude API Keys | Anthropic & AWS Bedrock API key management |
| Task Management | AI agent todo system |
| Rate Limit Handling | Smart API throttling & retry logic |
| Claude Chat History | Inspect & resume Claude Code sessions |
Note: This project is under active development. Contributions welcome!
uv install -e .Submit PRs to GitHub | Report issues in Issues
Apache 2.0 - See LICENSE
PyPI • Issues • Contribute