Easiest IDEs & Code Editors to self-host

Last reviewed Oct 11, 2026 · 682 words

Start with code-server. It tops this shortlist and is by far its most popular app (about 80,000 stars, MIT, maintained since 2019), it ships as one Docker image or one binary, and the directory estimates 512 MB of RAM for the editor itself. None of the three apps here earns the ≈5-minute badge, though. code-server and OpenVSCode Server publish images but leave the configuration to you, and Harlequin has no container path at all, so plan on an hour for the first install rather than a coffee break.

The three easy picks are not the same kind of tool

code-server and OpenVSCode Server answer the same question, and the code-server vs OpenVSCode Server comparison covers the differences. For a beginner the tiebreak is simple: code-server has roughly thirteen times the stars, the directory's estimate is half the RAM, and far more of the community answers you will find while debugging are written about it. Harlequin sits on this list because it is genuinely easy, but it is a local terminal program for querying DuckDB, SQLite, or PostgreSQL, not a server you reach from a browser. Install it on your laptop if that is the job; it does not replace a remote editor.

What "easy" still leaves you to do

Getting the container to start is the small part. Four jobs remain.

Persistence comes first. Mount a volume for the user's home directory so extensions, settings, and your projects survive a container upgrade. Without it, pulling a new image quietly resets you to a blank editor.

Access comes second, and this is where code-server differs from a recipe manager or a calendar: the integrated terminal runs commands as the container's user, so a leaked password is shell access. Keep it on a VPN, or put it behind a reverse proxy that adds its own login and passes websockets through. The remote access rundown lays out the options.

Backups are lighter than they look. Code belongs in git, pushed somewhere else, so what you actually back up is the home volume with settings and any unpushed work.

Accounts are the limitation. One instance means one user with one password, so a second person needs a second instance.

How much RAM does a self-hosted code editor need?

Budget for the toolchain, not the editor. The 512 MB figure is the directory's estimate for code-server idling with a file open. Language servers, a TypeScript watcher, or a Rust build in the terminal cost what they cost on a laptop, so 2 to 4 GB is a realistic floor for one person doing real work.

When to step up to a harder editor

Stay on code-server until you hit one of three walls. If you want each project to get a fresh, reproducible environment from a devcontainer file, DevPod (Medium) does that on infrastructure you already own. If your work is notebooks rather than source files, JupyterLab is the standard, and JupyterHub takes over when several people need one each. If a team needs workspaces with SSO, the category's buyer's guide covers the heavier platforms built for that.

What I'd do

Run code-server in Docker with a persistent home volume, reach it only over a VPN, and give the host 4 GB so the toolchain has room. Ignore OpenVSCode Server unless you have a specific reason to prefer Gitpod's build, and install Harlequin locally if SQL is your daily work. Move to DevPod or the Jupyter tools only when the job changes shape, not because the first editor felt too simple.