Run Claude Code in Containers

Run Claude Code in Containers | Generated by AI

Home 2026.02

Question: How to run Claude code (Claude Code interpreter/agent) in a container or cloud server?

Answer:

Claude Code (Anthropic’s terminal-based AI coding agent / code interpreter tool) is commonly run inside Docker containers for safety and isolation. This is the officially recommended approach in many cases because Claude Code can execute arbitrary code, install packages, run servers, etc.

Here are the main practical ways people run it in 2026 (local container or cloud / VPS):

1. Quickest & Recommended: Use Docker Sandbox (local or cloud VM)

Docker introduced built-in support for running coding agents like Claude Code in isolated sandboxes.

# Make sure you have recent Docker installed
docker sandbox run claude ~/my-project

Mounts your current folder
Runs Claude Code inside a secure, isolated environment
Very low setup effort

You can also pass a prompt directly:

docker sandbox run claude -- "refactor this API client to use async/await"

Works on Mac, Windows, Linux.

2. Run inside VS Code Dev Container (very popular & secure)

Most developer-oriented way:

Install Docker Desktop + VS Code + Dev Containers extension
Use Anthropic’s official template or community ones:
- https://github.com/anthropics/claude-code (look for .devcontainer folder)
- Many community forks / templates exist (search “claude code devcontainer”)
Open folder → “Reopen in Container”
Claude Code runs inside the container with strong isolation
You can give --dangerously-skip-permissions for fully autonomous mode

This setup is frequently used for long-running agents.

3. Plain Docker container (DIY / cloud VPS friendly)

Minimal example Dockerfile:

FROM ubuntu:24.04

RUN apt-get update && apt-get install -y curl git build-essential
RUN curl -fsSL https://claude.ai/install.sh | bash

# Optional: install common tools Claude Code likes
RUN apt-get install -y python3 python3-pip nodejs npm

WORKDIR /workspace

CMD ["claude"]

Build & run:

docker build -t claude-code .
docker run -it --rm \
  -v "$(pwd)":/workspace \
  -e ANTHROPIC_API_KEY=sk-ant-... \
  claude-code

On a cloud VPS (RunPod, Hostinger, AWS EC2, DigitalOcean, etc.):

SSH into the server
Install Docker
Use one of the above methods
Many people use tmux + claude or expose via ShellHub / Tailscale for remote access from phone/laptop

4. Cloud-specific / managed sandbox options

Cloudflare Sandboxes → run Claude Code via their sandbox SDK
E2B, Daytona, Vercel Sandbox, Fly Machines — offer managed secure containers for AI agents
RunPod — popular for GPU pods + Claude Code setups (good templates exist)
AWS ECS Fargate / Google Cloud Run — containerize Claude Code and run serverless / always-on

5. If you want fully local models (no Anthropic API)

Some people replace the Anthropic backend with local models via Docker Model Runner or Ollama + Anthropic-compatible endpoint:

ANTHROPIC_BASE_URL=http://localhost:12434 claude --model your-local-model ...

But most people still use real Claude models (Sonnet 3.7 / Opus 4 / etc.) via API key.

Security note: Always run Claude Code in a container — never directly on your main machine with dangerous flags, especially if giving it long-running / autonomous tasks.

Choose method depending on your goal:

Quick & local → docker sandbox run claude
Development workflow → VS Code Dev Container
24/7 cloud agent → VPS + Docker or managed sandbox (E2B, Cloudflare, RunPod)

References:

Back Donate