IDEs & Code Editors

Last reviewed Aug 26, 2026 · 441 words

Half of this category is the same editor in different wrappers. code-server, OpenVSCode Server, Eclipse Theia, and the workspaces inside Coder and Gitpod are all VS Code or VS Code-compatible, so the decision is rarely about the editor. It is about how much orchestration you want wrapped around it, and the range runs from a 512 MB single binary to an 8 GB Kubernetes install.

How to choose a self-hosted dev environment

Three axes split the field. First, solo or team: code-server is one user per instance, which is fine for a personal box and wrong for a company; Coder, Gitpod, and Eclipse Che exist to hand out workspaces to many people, with SSO and policies. Second, editor versus environment: code-server puts an editor on a machine you already manage, while DevPod and Coder provision reproducible environments (devcontainers for DevPod, Terraform templates for Coder) and let you bring VS Code or JetBrains to them. Third, the operations bill: code-server runs in 512 MB and installs in minutes, Coder is rated Hard with 1 GB just for the control plane, and self-hosted Gitpod wants Kubernetes, Helm, and a minimum of 8 GB. Do not buy team infrastructure for a party of one.

The extension catch nobody mentions up front

Microsoft licenses its extension Marketplace to official VS Code builds only, so code-server and OpenVSCode Server pull extensions from the Open VSX registry instead. Most things you use daily are there, but a few Microsoft-proprietary extensions are not, and the catalogue's own listing flags this: "some Marketplace extensions unavailable." Check your two or three must-have extensions exist on Open VSX before you commit an afternoon.

Where to start

For an individual developer, code-server is the obvious first install: 79,069 GitHub stars, MIT licensed, deployable as Docker or a single binary, and it makes an iPad or a borrowed laptop a real development machine. For a data or research workflow, JupyterLab is the notebook standard, and when a class or lab team needs it, JupyterHub (around since 2015) spawns and proxies a notebook server per user with pluggable authentication. For an engineering team that wants Codespaces without the metered billing, Coder defines workspaces in Terraform and stays editor-agnostic; budget real setup time, because its Hard difficulty rating is earned.

A first-timer should run code-server on whatever server they already own, put it behind a VPN or reverse proxy with authentication rather than an open port, and stop there. Graduate to Coder only when a second person needs a workspace.