Coder — Self-Hosted Cloud Development Environments

What it is

Coder is an open-source platform for provisioning self-hosted cloud development environments. Instead of coding on local machines with inconsistent setups, teams define environments as Terraform templates and spin them up on Kubernetes, Docker, AWS, GCP, or Azure.

Coder targets engineering teams that need reproducible, secure dev environments without vendor lock-in. Platform engineers use it to standardize tooling, while developers get instant workspaces with all dependencies pre-installed.

How it saves time or tokens

New developer onboarding drops from hours of local setup to minutes. A Terraform template defines the entire environment (OS, language runtimes, databases, IDE extensions), and Coder provisions it on demand. When something breaks, developers destroy and recreate the workspace instead of debugging local configuration drift.

Coder also centralizes secrets management. Database credentials and API keys live in the workspace template, not on individual laptops.

How to use

1. Install and start the Coder server:

curl -L https://coder.com/install.sh | sh
coder server

1. Open http://localhost:3000 and create your first template:

coder templates init
coder templates push my-template

1. Create a workspace from the template:

coder create my-workspace --template my-template

Example

A minimal Terraform template for a Docker-based workspace:

resource "coder_agent" "main" {
  os   = "linux"
  arch = "amd64"
}

resource "docker_container" "workspace" {
  image = "codercom/enterprise-base:ubuntu"
  name  = "coder-${data.coder_workspace.me.name}"
  command = ["sh", "-c", coder_agent.main.init_script]
  env = ["CODER_AGENT_TOKEN=${coder_agent.main.token}"]
}

This creates a Linux container with the Coder agent pre-configured. Developers connect via VS Code Remote, JetBrains Gateway, or the built-in web terminal.

Related on TokRepo

• AI tools for DevOps -- Infrastructure automation and developer platform tools
• Self-hosted tools -- Self-hosted alternatives for development workflows

Common pitfalls

• Over-provisioning workspace resources. Start with small CPU/memory limits and let developers request more. Cloud compute costs scale linearly with workspace count.
• Ignoring template versioning. When you push a new template version, existing workspaces do not auto-update. Communicate breaking changes to your team.
• Skipping the idle timeout configuration. Workspaces left running overnight waste compute. Set auto-stop after 30-60 minutes of inactivity.