Install your own command-line tools

Install Homebrew and miseCopy Anchor Link

Open a terminal in your workspace.

Install Homebrew with its setup script:

NONINTERACTIVE=1 /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

Homebrew installs to /home/linuxbrew/.linuxbrew. Add it to your shell so the brew command is available:

echo 'eval "$(/home/linuxbrew/.linuxbrew/bin/brew shellenv)"' >> ~/.bashrc
eval "$(/home/linuxbrew/.linuxbrew/bin/brew shellenv)"

Confirm Homebrew is available before you continue:

brew --version

Install mise with its setup script:

curl -fsSL https://mise.run | sh

mise installs to ~/.local/bin/mise, inside your home directory. Activate it so every new shell loads it:

echo 'eval "$(~/.local/bin/mise activate bash)"' >> ~/.bashrc

Activation only takes effect in shells that start after this change, so open a new terminal (or run source ~/.bashrc) before you use mise. Until you do, mise doctor reports that mise isn't activated, which is expected at this point. For other shells, refer to Activate mise.

Open a new terminal so both the Homebrew and mise changes take effect, then confirm each manager runs:

brew --version
mise --version

Install a tool with each managerCopy Anchor Link

Install ripgrep with Homebrew:

brew install ripgrep

Install bat with mise:

mise use -g bat

Confirm both tools run:

rg --version
bat --version

Both work. So far, the two package managers look interchangeable.

Restart the workspace and compareCopy Anchor Link

Restart the workspace. The restart rebuilds the container from the image and keeps only your home directory.

When you reconnect, your shell prints an error before you run anything:

bash: /home/linuxbrew/.linuxbrew/bin/brew: No such file or directory

That's the first sign something changed. Your .bashrc still tries to load Homebrew, but the restart removed it. Check each tool to see what survived.

bat, installed with mise, still works:

bat --version

bat 0.26.1

rg, installed with Homebrew, is gone:

rg --version

bash: rg: command not found

So is brew itself:

brew --version

bash: brew: command not found

mise installed bat under /home/coder, which persists, so bat survived. Homebrew installed ripgrep to /home/linuxbrew, outside /home/coder, so the rebuild discarded Homebrew and every formula you installed with it. The brew shellenv line stayed in your .bashrc because it lives in /home/coder, which is why your shell still tries to load the missing brew and prints the error above.

Make Homebrew survive restartsCopy Anchor Link

To make Homebrew survive a restart, you'll edit the template and add a persistent volume. The volume backs /home/linuxbrew, the prefix where Homebrew installs, so Homebrew and its formulae stay between restarts.

In main.tf, add a volume for Homebrew's directory next to the existing home_volume:

resource "docker_volume" "homebrew_volume" {
  name = "coder-${data.coder_workspace.me.id}-homebrew"
  lifecycle {
    ignore_changes = all
  }
}

Then mount it in the docker_container "workspace" resource, alongside the block that mounts /home/coder:

  volumes {
    container_path = "/home/linuxbrew"
    volume_name    = docker_volume.homebrew_volume.name
    read_only      = false
  }

Publish the change and restart the workspace:

The restart gives you a persistent but empty /home/linuxbrew. Your earlier Homebrew install is gone, so install it once more. This time it lands on the volume:

NONINTERACTIVE=1 /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

The brew shellenv line is still in your .bashrc from Step 1, so open a new terminal to load it, then reinstall ripgrep:

brew install ripgrep

Restart the workspace once more and reconnect. This time the startup error is gone, and both tools report their versions:

rg --version
brew --version

Homebrew now persists because /home/linuxbrew lives on its own volume, the same way mise's tools persist because they live in /home/coder.

Install your tools in every new workspaceCopy Anchor Link

Steps 1 through 4 set up your tools in this workspace. To give everyone who uses the template the same tools, install them from the template so every new workspace starts with them.

The template already installs languages on every start with the install_languages script. You'll add a second script that installs your command-line tools the same way.

mise is the lighter choice here. It installs prebuilt binaries into /home/coder, which already persists, so the first start stays quick and later starts reuse the tools.

In main.tf, add a script next to the existing install_languages resource:

resource "coder_script" "install_tools" {
  agent_id     = coder_agent.main.id
  display_name = "Install Tools"
  icon         = "/icon/terminal.svg"
  run_on_start = true
  script       = <<-EOT
    #!/usr/bin/env bash
    set -e

    # Install mise on first start. It lives in /home/coder, so later starts reuse it.
    if [ ! -x "$HOME/.local/bin/mise" ]; then
      curl -fsSL https://mise.run | sh
    fi
    export PATH="$HOME/.local/bin:$HOME/.local/share/mise/shims:$PATH"

    # Install the tools for everyone who uses the template, tracking the latest release.
    mise use -g ripgrep@latest bat@latest

    # Load mise in new interactive shells so the tools are on PATH.
    if ! grep -qs 'mise activate' "$HOME/.bashrc"; then
      echo 'eval "$(mise activate bash)"' >> "$HOME/.bashrc"
    fi
  EOT
}

mise use -g ripgrep@latest bat@latest writes the tools to mise's global config at ~/.config/mise/config.toml and installs them, so every workspace from the template resolves the same versions.

Publish the change and apply it to your workspace:

When the workspace is back, confirm both tools run without installing anything by hand:

rg --version
bat --version

The script runs on every start, so every new workspace from the template now ships with ripgrep and bat.