claudebox

A runtime harness for Claude Code — the agentic coding CLI from Anthropic — running in a fully isolated Docker container with every dev tool pre-installed, passwordless sudo, docker-in-docker support, and --permission-mode bypassPermissions enabled by default.

v2.0.0 — rebased on psyb0t/aicodebox. claudebox is now a thin child image of the shared aicodebox base (same pattern as psyb0t/pibox). Every mode surface (API / Telegram / Cron / MCP) is inherited from the base and stays in lockstep with future base fixes. See CHANGELOG.md for the full migration guide (endpoint shape changes, env-var namespace, path renames — all mitigated by aliases + symlinks so existing configs keep working).

Runtime hardening (recommended docker run flags):

• --cap-drop=ALL --cap-add=NET_BIND_SERVICE — drop every Linux capability, add back only bind-below-1024 if you actually need it.
• --security-opt no-new-privileges:true — block setuid privilege escalation inside the container.
• --memory=2g --cpus=2 --pids-limit=512 — cap runtime resource use so a runaway process can't starve the host.
• --read-only --tmpfs /tmp:rw,noexec,nosuid (only if you don't use /workspace for writes — otherwise skip). The container drops from root to aicode (UID 1000) at boot via setpriv in the base entrypoint, so the process running your code is never root even without --user.

claudebox wraps Claude Code with several distinct interfaces:

• Interactive CLI — a drop-in replacement for the native claude command, with persistent containers and automatic session resumption across runs
• Programmatic CLI — non-interactive mode for scripts, CI/CD pipelines, and automation; pass a prompt, get structured output, pipe it wherever you need
• HTTP API server — a full REST API with workspace management, file operations, structured output formats, and workspace isolation for multi-tenant deployments
• OpenAI-compatible endpoint — a chat/completions adapter that lets LiteLLM, OpenAI SDKs, and any OpenAI-compatible client talk to Claude Code, complete with streaming SSE, multi-turn conversations, and multimodal image handling
• MCP server — a Model Context Protocol endpoint over streamable HTTP so other AI agents and tools (Claude Desktop, other Claude Code instances, etc.) can use Claude Code as a tool
• Telegram bot — a conversational interface with per-chat workspaces, configurable models and effort levels, file sharing, shell access, and group chat support
• Cron scheduler — yaml-defined Claude jobs running on cron schedules with per-job activity history, sub-minute resolution, and overlap protection

Beyond just running Claude Code in Docker, claudebox adds skill injection (auto-load SKILL.md files into every session), init hooks, custom script directories, structured JSON logging, and a workspace management layer that handles multi-tenant isolation with automatic busy/idle tracking.

Renamed from docker-claude-code: This project was previously called docker-claude-code with the Docker image at psyb0t/claude-code. Starting with v1.0.0, it is claudebox — the Docker image is now psyb0t/claudebox, the default binary name is claudebox, the GitHub repository is psyb0t/docker-claudebox, and the SSH key directory defaults to ~/.ssh/claudebox. If you were using the old names, update your image references, wrapper scripts, and SSH paths accordingly.

Table of Contents

• Requirements
• Quick Start
• Image Variants
• What's Inside (Full Image)
• Authentication
• Modes
  • Interactive mode
  • Programmatic mode
  • API mode
  • Telegram mode
  • Cron mode
  • MCP mode
• Configuration
• Agent integrations
• Gotchas
• License

Requirements

Docker installed and running. That's it.

Quick Start

One-liner install

The install script pulls the Docker image, generates SSH keys for git operations inside the container, downloads the wrapper script, and installs it as a command on your system.

# minimal image — default; Claude installs what it needs on the fly
curl -fsSL https://raw.githubusercontent.com/psyb0t/docker-claudebox/master/install.sh | bash

# full image — every dev tool pre-installed (Go, Python, kubectl, terraform, ...)
export CLAUDEBOX_FULL=1 && curl -fsSL https://raw.githubusercontent.com/psyb0t/docker-claudebox/master/install.sh | bash

# custom binary name (e.g. if you want to call it 'claude' instead of 'claudebox')
curl -fsSL https://raw.githubusercontent.com/psyb0t/docker-claudebox/master/install.sh | bash -s -- claude
# or: export CLAUDEBOX_BIN_NAME=claude && curl -fsSL .../install.sh | bash

v2 note: the variant naming flipped in v2. latest is now the minimal image (was the full image pre-v2); latest-full is the toolchain-loaded variant (was latest pre-v2). The CLAUDEBOX_MINIMAL=1 opt-in from v1 is now a no-op — you already get minimal by default. Set CLAUDEBOX_FULL=1 to opt into the toolchain image. Installing with CLAUDEBOX_FULL=1 (as above) bakes the choice into the installed wrapper, so the full variant sticks for every run — you don't need to keep the env var set afterward.

Heads up on env vars: VAR=x curl … | bash does not set VAR for the install script — bash semantics attach the var to curl only. Always export the var first (or put it on the bash side of the pipe).

Manual setup

If you prefer not to pipe scripts to bash:

# 1. create the data directory
mkdir -p ~/.claude

# 2. create SSH keys for git operations inside the container
mkdir -p "$HOME/.ssh/claudebox"
ssh-keygen -t ed25519 -C "claude@claude.ai" -f "$HOME/.ssh/claudebox/id_ed25519" -N ""
# then add the public key to GitHub/GitLab/wherever you push code

# 3. pull the image
docker pull psyb0t/claudebox:latest        # minimal (default)
# or: docker pull psyb0t/claudebox:latest-full   # toolchain-loaded variant

# 4. grab the wrapper script and install it
# see install.sh for exactly how the wrapper is set up

Image Variants

psyb0t/claudebox:latest (minimal, default)

The default v2 image. Just enough to run Claude Code on top of the aicodebox base: Ubuntu 24.04, git/curl/wget/jq, Node.js 24 LTS + npm, Python 3.14 + uv, Docker CE. Claude has passwordless sudo, so it will install whatever else it needs on the fly via apt-get, pip, npm, etc. Smaller image, faster pull, first run may take longer while Claude sorts out its own tooling.

Claude Code is installed on first run, not baked into the image. Anthropic's Claude Code CLI is proprietary and can't be redistributed, so the image ships only the pinned version (CLAUDEBOX_CLAUDE_VERSION, default set at build) and the entrypoint runs npm install -g @anthropic-ai/claude-code@<version> from npm the first time a fresh container starts. This means the published image redistributes none of Anthropic's software, and each container pulls Claude Code straight from npm. First container start needs network and takes a few extra seconds; warm restarts skip it. To pin a different version, set CLAUDEBOX_CLAUDE_VERSION at docker run.

curl -fsSL https://raw.githubusercontent.com/psyb0t/docker-claudebox/master/install.sh | bash

Use /aicodebox-init.d/*.sh hooks (see Init Hooks) to pre-install your tools on first container create so Claude doesn't burn tokens figuring out package management.

psyb0t/claudebox:latest-full (toolchain-loaded)

Everything pre-installed. Layered on top of the minimal image: Go 1.26.7, Python 3.14.7 via pyenv, Node.js dev tools, C/C++ toolchain, terraform, kubectl, helm, gh, database clients (sqlite/postgres/mysql/redis), editors (vim/nano/htop/tmux), linters + formatters (flake8/black/isort/pyright/mypy/ruff/eslint/prettier/gofumpt/…). Larger image but Claude wakes up ready.

export CLAUDEBOX_FULL=1 && curl -fsSL https://raw.githubusercontent.com/psyb0t/docker-claudebox/master/install.sh | bash

Comparison