Container Use provides sandboxed development environments for coding agents. It leverages Dagger for containerized workflows and Git worktrees for branching.

Key features include:

• Sandboxed Environments: Each agent operates in a fresh container within its own Git branch, allowing multiple agents to run in parallel without conflicts.
• Real-time Visibility: Provides complete command history and logs of agent actions.
• Direct Intervention: Allows users to drop into an agent's terminal to inspect state or take manual control.
• Environment Control: Supports standard Git workflows; users can review an agent's work by running git checkout <branch_name>.

An environment is an isolated, containerized development workspace that combines Docker containers with Git branches. It provides agents with safe, persistent workspaces where all changes are automatically tracked and versioned.

Each environment consists of:

• Git Branch: A dedicated branch tracking all changes and history.
• Container: A Dagger container containing your code and dependencies.
• History: Versioned snapshots of container state changes appended to the branch as Git notes.
• Configuration: Base image, setup commands, secrets, and instructions that can be checked into the source repository.

To use container-use with Claude Code, add it as an MCP server using the stdio transport. You can also optionally add agent rules to your CLAUDE.md to improve agent behavior.

Note: The container-use command can be shortened to cu.

# Add Container Use MCP server
cd /path/to/repository
claude mcp add container-use -- container-use stdio

# Add agent rules (optional)
curl https://raw.githubusercontent.com/dagger/container-use/main/rules/agent.md >> CLAUDE.md

Before installing, ensure you have Docker and Git installed. You can install Container Use using Homebrew (macOS), a shell script (all platforms), or by building from source.

# Homebrew (macOS)
brew install dagger/tap/container-use

# Shell Script (All Platforms)
curl -fsSL https://raw.githubusercontent.com/dagger/container-use/main/install.sh | bash

# Build from Source
git clone https://github.com/dagger/container-use.git
cd container-use
go build -o container-use ./cmd/container-use
sudo mv container-use /usr/local/bin/

# Verify installation
container-use version

Configure the Container Use MCP server in an opencode.json file and optionally add an AGENTS.md file for best results.

# Configure opencode.json
{
  "$schema": "http://opencode.ai/config.json",
  "mcp": {
    "container-use": {
      "type": "local",
      "command": ["container-use", "stdio"],
      "enabled": true
    }
  }
}

# Add AGENTS.md (Optional)
curl https://raw.githubusercontent.com/dagger/container-use/main/rules/agent.md >> AGENTS.md

Add the container-use MCP server configuration to ~/.config/amp/settings.json and optionally add agent rules to your project root.

# Add to ~/.config/amp/settings.json
{
    "amp.mcpServers": {
        "container-use": {
            "command": "container-use",
            "args": ["stdio"]
        }
    },
    "amp.permissions": [
      {"tool": "mcp__container-use__*", "action": "allow"}
    ],
    "amp.dangerouslyAllowAll": false,
    "amp.updates.autoUpdate.enabled": true
}

# Add Agent Rules
curl https://raw.githubusercontent.com/dagger/container-use/main/rules/agent.md >> AGENT.md

Configure the MCP server in your VSCode settings and optionally add Copilot instructions to your project.

# Update VSCode settings
"mcp": {
  "servers": {
    "container-use": {
      "type": "stdio",
      "command": "container-use",
      "args": ["stdio"]
    }
  }
}

# Add Copilot Instructions (Optional)
curl --create-dirs -o .github/copilot-instructions.md https://raw.githubusercontent.com/dagger/container-use/main/rules/agent.md

Agents may modify their environment during work (e.g., installing a new tool). These changes are ephemeral to that specific environment unless you view them or import them into your defaults.

• Use container-use config show <env> to inspect an agent's modified environment.
• Use container-use config import <env> to adopt those changes as your new project defaults.

# View your default configuration
container-use config show

# View an agent's modified configuration (e.g., environment named 'fancy-mallard')
container-use config show fancy-mallard

# Output configuration as JSON
container-use config show --json

# Import agent changes as new defaults
container-use config import fancy-mallard

You can integrate Container Use with Goose using a configuration file, interactive setup, or the Goose Desktop URL.

# Method 1: Add to ~/.config/goose/config.yaml
extensions:
  container-use:
    name: container-use
    type: stdio
    enabled: true
    cmd: container-use
    args:
      - stdio
    envs: {}

# Method 2: Interactive Setup
goose configure
# Then add a command line extension with 'container-use stdio' as the command

Method 3: Goose Desktop

Paste this URL into your browser:

goose://extension?cmd=container-use&arg=stdio&id=container-use&name=container%20use&description=use%20containers%20with%20dagger%20and%20git%20for%20isolated%20environments