dive

A tool for exploring a Docker image, layer contents, and discovering ways to shrink the size of your Docker/OCI image.

To analyze a Docker image simply run dive with an image tag/id/digest:

dive <your-image-tag>

or you can dive with Docker directly:

alias dive="docker run -ti --rm  -v /var/run/docker.sock:/var/run/docker.sock docker.io/wagoodman/dive"
dive <your-image-tag>

# for example
dive nginx:latest

or if you want to build your image then jump straight into analyzing it:

dive build -t <some-tag> .

Building on macOS (supporting only the Docker container engine):

docker run --rm -it \
      -v /var/run/docker.sock:/var/run/docker.sock \
      -v  "$(pwd)":"$(pwd)" \
      -w "$(pwd)" \
      -v "$HOME/.dive.yaml":"$HOME/.dive.yaml" \
      docker.io/wagoodman/dive:latest build -t <some-tag> .

Additionally you can run this in your CI pipeline to ensure you're keeping wasted space to a minimum (this skips the UI):

CI=true dive <your-image>

This is beta quality! Feel free to submit an issue if you want a new feature or find a bug :)

Model Context Protocol (MCP) Server

dive can act as an MCP server, allowing AI agents (like Claude Desktop or IDE extensions) to programmatically analyze images, inspect layers, and perform deep diffing between image states.

Starting the Server

Stdio Transport (Default for local agents):

dive mcp

HTTP with SSE Transport (For remote or containerized access):

dive mcp --transport sse --host localhost --port 8080

Available Tools

• analyze_image(image, source): Returns a structured JSON summary of efficiency metrics and layer metadata.
• get_wasted_space(image, source): Returns a JSON list of the top inefficient files (duplicated or moved).
• inspect_layer(image, layer_index, path, source): Lists files and directories within a specific layer and path.
• diff_layers(image, base_layer_index, target_layer_index, source): Returns a detailed JSON delta (added/modified/removed) between any two layers.

Dynamic Resources

You can reference analysis results as URIs:

• dive://image/{name}/summary: Get the latest analysis summary as JSON.
• dive://image/{name}/efficiency: Get just the efficiency score and wasted bytes.

Configuration for Claude Desktop

Add the following to your claude_desktop_config.json:

For Stdio:

{
  "mcpServers": {
    "dive": {
      "command": "/path/to/dive",
      "args": ["mcp", "--quiet"]
    }
  }
}

For HTTP/SSE:

{
  "mcpServers": {
    "dive": {
      "url": "http://localhost:8080/sse"
    }
  }
}

Security & Performance

• Sandbox: Use --mcp-sandbox /path/to/archives to restrict the server's access to local tarballs.
• Caching: The server uses an LRU cache for analysis results. Configure it with --mcp-cache-size and --mcp-cache-ttl.

Persistent Configuration

You can also configure the MCP server in your .dive.yaml file:

mcp:
  # The transport to use: stdio or sse
  transport: sse
  host: 0.0.0.0
  port: 8080
  
  # Security sandbox for local archives
  sandbox: /path/to/archives
  
  # Performance tuning
  cache-size: 20
  cache-ttl: 24h

Basic Features

Show Docker image contents broken down by layer

As you select a layer on the left, you are shown the contents of that layer combined with all previous layers on the right. Also, you can fully explore the file tree with the arrow keys.

Indicate what's changed in each layer

Files that have changed, been modified, added, or removed are indicated in the file tree. This can be adjusted to show changes for a specific layer, or aggregated changes up to this layer.

Estimate "image efficiency"

The lower left pane shows basic layer info and an experimental metric that will guess how much wasted space your image contains. This might be from duplicating files across layers, moving files across layers, or not fully removing files. Both a percentage "score" and total wasted file space is provided.

Quick build/analysis cycles

You can build a Docker image and do an immediate analysis with one command: dive build -t some-tag .

You only need to replace your docker build command with the same dive build command.

CI Integration

Analyze an image and get a pass/fail result based on the image efficiency and wasted space. Simply set CI=true in the environment when invoking any valid dive command.

Multiple Image Sources and Container Engines Supported

With the --source option, you can select where to fetch the container image from:

dive <your-image> --source <source>

or

dive <source>://<your-image>

With valid source options as such:

• docker: Docker engine (the default option)
• docker-archive: A Docker Tar Archive from disk
• podman: Podman engine (linux only)

Installation

Ubuntu/Debian

Using debs:

DIVE_VERSION=$(curl -sL "https://api.github.com/repos/wagoodman/dive/releases/latest" | grep '"tag_name":' | sed -E 's/.*"v([^"]+)".*/\1/')
curl -fOL "https://github.com/wagoodman/dive/releases/download/v${DIVE_VERSION}/dive_${DIVE_VERSION}_linux_amd64.deb"
sudo apt install ./dive_${DIVE_VERSION}_linux_amd64.deb

Using snap:

sudo snap install docker
sudo snap install dive
sudo snap connect dive:docker-executables docker:docker-executables
sudo snap connect dive:docker-daemon docker:docker-daemon

Caution

The Snap method is not recommended if you installed Docker via apt-get, since it might break your existing Docker daemon.

See also: https://github.com/wagoodman/dive/issues/546

RHEL/Centos

DIVE_VERSION=$(curl -sL "https://api.github.com/repos/wagoodman/dive/releases/latest" | grep '"tag_name":' | sed -E 's/.*"v([^"]+)".*/\1/')
curl -fOL "https://github.com/wagoodman/dive/releases/download/v${DIVE_VERSION}/dive_${DIVE_VERSION}_linux_amd64.rpm"
rpm -i dive_${DIVE_VERSION}_linux_amd64.rpm

Arch Linux

Available in the extra repository and can be installed via pacman:

pacman -S dive

Mac

If you use Homebrew:

brew install dive

If you use MacPorts:

sudo port install dive

Or download the latest Darwin build from the releases page.

Windows

If you use Chocolatey

choco install dive

If you use scoop

scoop install main/dive

If you use winget:

winget install --id wagoodman.dive

Or download the latest Windows build from the releases page.

Go tools Requires Go version 1.10 or higher.

go install github.com/wagoodman/dive@latest

Note: installing in this way you will not see a proper version when running dive -v.

Nix/NixOS

On NixOS:

nix-env -iA nixos.dive

On non-NixOS (Linux, Mac)

nix-env -iA nixpkgs.dive

X-CMD

x-cmd is a toolbox for Posix Shell, offering a lightweight package manager built using shell and awk.

x env use dive

Docker

docker pull docker.io/wagoodman/dive
# or alternatively
docker pull ghcr.io/wagoodman/dive

When running you'll need to include the Docker socket file:

docker run --rm -it \
    -v /var/run/docker.sock:/var/run/docker.sock \
    docker.io/wagoodman/dive:latest <dive arguments...>

Docker for Windows (showing PowerShell compatible line breaks; collapse to a single line for Command Prompt compatibility)

docker run --rm -it `
    -v /var/run/docker.sock:/var/run/docker.sock `
    docker.io/wagoodman/dive:latest <dive arguments...>

Note: depending on the version of docker you are running locally you may need to specify the docker API version as an environment variable:

   DOCKER_API_VERSION=1.37 dive ...

or if you are running with a docker image:

docker run --rm -it \
    -v /var/run/docker.sock:/var/run/docker.sock \
    -e DOCKER_API_VERSION=1.37 \
    docker.io/wagoodman/dive:latest <dive arguments...>

if you are using an alternative runtime (Colima etc) then you may need to specify the docker host as an environment variable in order to pull local images:

   export DOCKER_HOST=$(docker context inspect -f '{{ .Endpoints.docker.Host }}')

CI Integration

When running dive with the environment variable CI=true then the dive UI will be bypassed and will instead analyze your docker image, giving it a pass/fail indication via return code. Currently there are three metrics supported via a .dive-ci file that you can put at the root of your repo:

rules:
  # If the efficiency is measured below X%, mark as failed.
  # Expressed as a ratio between 0-1.
  lowestEfficiency: 0.95

  # If the amount of wasted space is at least X or larger than X, mark as failed.
  # Expressed in B, KB, MB, and GB.
  highestWastedBytes: 20MB

  # If the amount of wasted space makes up for X% or more of the image, mark as failed.
  # Note: the base image layer is NOT included in the total image size.
  # Expressed as a ratio between 0-1; fails if the threshold is met or crossed.
  highestUserWastedPercent: 0.20

You can override the CI config path with the --ci-config option.

KeyBindings

Key Binding	Description
Ctrl + C or Q	Exit
Tab	Switch between the layer and filetree views
Ctrl + F	Filter files
ESC	Close filter files
PageUp or U	Scroll up a page
PageDown or D	Scroll down a page
Up or K	Move up one line within a page
Down or J	Move down one line within a page
Ctrl + A	Layer view: see aggregated image modifications
Ctrl + L	Layer view: see current layer modifications
Space	Filetree view: collapse/uncollapse a directory
Ctrl + Space	Filetree view: collapse/uncollapse all directories
Ctrl + A	Filetree view: show/hide added files
Ctrl + R	Filetree view: show/hide removed files
Ctrl + M	Filetree view: show/hide modified files
Ctrl + U	Filetree view: show/hide unmodified files
Ctrl + B	Filetree view: show/hide file attributes
PageUp or U	Filetree view: scroll up a page
PageDown or D	Filetree view: scroll down a page

UI Configuration

No configuration is necessary, however, you can create a config file and override values:

# supported options are "docker" and "podman"
container-engine: docker
# continue with analysis even if there are errors parsing the image archive
ignore-errors: false
log:
  enabled: true
  path: ./dive.log
  level: info

# Note: you can specify multiple bindings by separating values with a comma.
# Note: UI hinting is derived from the first binding
keybinding:
  # Global bindings
  quit: ctrl+c
  toggle-view: tab
  filter-files: ctrl+f, ctrl+slash
  close-filter-files: esc
  up: up,k
  down: down,j
  left: left,h
  right: right,l

  # Layer view specific bindings
  compare-all: ctrl+a
  compare-layer: ctrl+l

  # File view specific bindings
  toggle-collapse-dir: space
  toggle-collapse-all-dir: ctrl+space
  toggle-added-files: ctrl+a
  toggle-removed-files: ctrl+r
  toggle-modified-files: ctrl+m
  toggle-unmodified-files: ctrl+u
  toggle-filetree-attributes: ctrl+b
  page-up: pgup,u
  page-down: pgdn,d

diff:
  # You can change the default files shown in the filetree (right pane). All diff types are shown by default.
  hide:
    - added
    - removed
    - modified
    - unmodified

filetree:
  # The default directory-collapse state
  collapse-dir: false

  # The percentage of screen width the filetree should take on the screen (must be >0 and <1)
  pane-width: 0.5

  # Show the file attributes next to the filetree
  show-attributes: true

layer:
  # Enable showing all changes from this layer and every previous layer
  show-aggregated-changes: false

dive will search for configs in the following locations:

• $XDG_CONFIG_HOME/dive/*.yaml
• $XDG_CONFIG_DIRS/dive/*.yaml
• ~/.config/dive/*.yaml
• ~/.dive.yaml

.yml can be used instead of .yaml if desired.