Container Use

At a glance

Container Use comes from Dagger, the team behind the CI engine. Instead of letting an agent edit your working copy directly, each task runs in a fresh container on its own git branch. You can run several agents side by side, read the full log of commands each one ran, open a terminal inside an agent's environment when it gets stuck, and review the result with an ordinary git checkout.

It is a command line tool that also speaks MCP, and Dagger still marks it experimental. It needs Docker or another container runtime that Dagger supports.

Made by

Dagger (dagger/container-use)

GitHub stars

4.1K

Last commit

21 Sept 2026

Tools

15

Runs on

Your computer

Our check: It needs something our Linux test machine does not have (an account, a desktop app, a Mac or a running service), so on 27 Oct 2026 we read its tools from the source code instead of running this config.

What's in the pod: its tools

We found 15 tools in its source code on 27 Oct 2026. 0 only read. The rest can change things, so keep your app's "ask before each tool call" setting on for them.

How we tested it: Read from the Go source on the main branch. Container Use ships as a compiled binary that needs Docker and Dagger, which our test machine could not run.

Grouped and described by us. The README on GitHub lists the inputs each tool takes.

Set it up

1. Install the CLI: brew install dagger/tap/container-use on macOS, or the install script from the repository on Linux.
2. In Claude Code, from your repository: claude mcp add container-use -- container-use stdio. Other clients use the config below.
3. Optional: append the project's agent rules file to your CLAUDE.md or AGENTS.md so the agent always works inside an environment.

Pick your app for the exact snippet and where it goes.

{
  "mcpServers": {
    "container-use": {
      "command": "container-use",
      "args": [
        "stdio"
      ]
    }
  }
}

claude mcp add-json container-use '{"command":"container-use","args":["stdio"]}'

{
  "mcpServers": {
    "container-use": {
      "command": "container-use",
      "args": [
        "stdio"
      ]
    }
  }
}

{
  "servers": {
    "container-use": {
      "type": "stdio",
      "command": "container-use",
      "args": [
        "stdio"
      ]
    }
  }
}

{
  "mcpServers": {
    "container-use": {
      "command": "container-use",
      "args": [
        "stdio"
      ]
    }
  }
}

The shorter cu command works too: cu stdio.

New to this? See the setup guides for Claude Code, Cursor, Claude Desktop and ChatGPT.

Settings

Names are exactly as the project spells them. Required ones are marked.

What it can touch

Safer setup: That is the point of the tool: the agent works in containers on separate branches, and nothing reaches your branch until you run cu apply or cu merge.

Before you start

• macOS or Linux (Windows is not supported)
• Docker (or another runtime Dagger supports)
• git

If it does not work

Creating an environment fails straight away

Docker (or another runtime Dagger supports) must be running. Container Use reports Docker errors with the fix.

The agent edits your files directly instead

Add the project's agent rules file to CLAUDE.md or AGENTS.md, and turn off the client's own file and shell tools for that session.

Every environment reinstalls the same packages

Move the installs into cu config setup-command so they are part of the environment's build.

A service the agent started is not reachable

Ask it to expose the port in environment_run_cmd; the result gives the address to open.

Peas in a pod: similar servers

How Container Use lines up against other agent helpers and desktop control servers on the shelf.

Questions

Something wrong on this page? Tell us.