Coder has open-sourced a new remote development platform 🥳 Check it out at coder/coder on GitHub.



This document is meant to serve current and future maintainers of code-server, as well as share our workflow for maintaining the project.


Current maintainers:

  • @code-asher
  • @jsjoeio

Occasionally, other Coder employees may step in time to time to assist with code-server.


To onboard a new maintainer to the project, please make sure to do the following:


Very similar to Onboarding but Remove maintainer from all teams and revoke access. Please also do the following:

  • Write farewell post via Discussion (see example)


The workflow used by code-server maintainers aims to be easy to understood by the community and easy enough for new maintainers to jump in and start contributing on day one.


We operate mainly using milestones. This was heavily inspired by our friends over at vscode.

Here are the milestones we use and how we use them:

  • "Backlog" -> Work not yet planned for a specific release.
  • "On Deck" -> Work under consideration for upcoming milestones.
  • "Backlog Candidates" -> Work that is not yet accepted for the backlog. We wait for the community to weigh in.
  • "<0.0.0>" -> Work to be done for a specific version.

With this flow, any un-assigned issues are essentially in triage state. Once triaged, issues are either "Backlog" or "Backlog Candidates". They will eventually move to "On Deck" (or be closed). Lastly, they will end up on a version milestone where they will be worked on.


We use the following process for triaging GitHub issues:

  1. Create an issue
  2. Add appropriate labels to the issue (including "needs-investigation" if we should look into it further)
  3. Add the issue to a milestone
    1. If it should be fixed soon, add to version milestone or "On Deck"
    2. If not urgent, add to "Backlog"
    3. Otherwise, add to "Backlog Candidate" for future consideration

Project boards

We use project boards for projects or goals that span multiple milestones.

Think of this as a place to put miscellaneous things (like testing, clean up stuff, etc). As a maintainer, random tasks may come up here and there. The project boards give you places to add temporary notes before opening a new issue. Given that our release milestones function off of issues, we believe tasks should have dedicated issues.

Project boards also give us a way to separate the issue triage from bigger-picture, long-term work.



The code-server project follows traditional semantic versioning, with the objective of minimizing major changes that break backward compatibility. We increment the patch level for all releases, except when the upstream Visual Studio Code project increments its minor version or we change the plugin API in a backward-compatible manner. In those cases, we increment the minor version rather than the patch level.

Pull requests

Ideally, every PR should fix an issue. If it doesn't, make sure it's associated with a version milestone.

If a PR does fix an issue, don't add it to the version milestone. Otherwise, the version milestone will have duplicate information: the issue and the PR fixing the issue.

Merge strategies

For most things, we recommend the squash and merge strategy. There may be times where creating a merge commit makes sense as well. Use your best judgment. If you're unsure, you can always discuss in the PR with the team.


To save time when creating a new release for code-server, we keep a running changelog at

If either the author or reviewer of a PR believes the change should be mentioned in the changelog, then it should be added.

If there is not a Next Version when you modify, please add it using the template you see near the top of the changelog.

When writing your changelog item, ask yourself:

  1. How do these changes affect code-server users?
  2. What actions do they need to take (if any)?

If you need inspiration, we suggest looking at the Emacs changelog.


With each release, we rotate the role of release manager to ensure every maintainer goes through the process. This helps us keep documentation up-to-date and encourages us to continually review and improve the flow.

If you're the current release manager, follow these steps:

  1. Create a release issue
  2. Fill out checklist
  3. Publish the release
  4. After release is published, close release milestone

Publishing a release

  1. Create a new branch called v0.0.0 (replace 0s with actual version aka v4.4.0)
  2. Run yarn release:prep and type in the new version (e.g., 3.8.1)
  3. GitHub Actions will generate the npm-package, release-packages and release-images artifacts. You do not have to wait for this step to complete before proceeding.
  4. Run yarn release:github-draft to create a GitHub draft release from the template with the updated version.
  5. Summarize the major changes in the release notes and link to the relevant issues.
  6. Change the @ to target the version branch. Example: v3.9.0 @ Target: v3.9.0
  7. Wait for the npm-package, release-packages and release-images artifacts to build.
  8. Run yarn release:github-assets to download the release-packages artifact. They will upload them to the draft release.
  9. Run some basic sanity tests on one of the released packages (pay special attention to making sure the terminal works).
  10. Publish the release and merge the PR. CI will automatically grab the artifacts, publish the NPM package from npm-package, and publish the Docker Hub image from release-images.
  11. Update the AUR package. Instructions for updating the AUR package are at coder/code-server-aur.
  12. Wait for the npm package to be published.


We publish to AUR as a package here. This process is manual and can be done by following the steps in this repo.


We publish code-server as a Docker image here, tagging it both with the version and latest.

This is currently automated with the release process.


We publish code-server on Homebrew here.

This is currently automated with the release process (but may fail occasionally). If it does, run this locally:

# Replace VERSION with version
brew bump-formula-pr --version="${VERSION}" code-server --no-browse --no-audit


We publish code-server as a npm package here.

This is currently automated with the release process.

Syncing with upstream Code

Refer to the contributing docs for information on how to update Code within code-server.


Our testing structure is laid out under our Contributing docs.

We hope to eventually hit 100% test coverage with our unit tests, and maybe one day our scripts (coverage not tracked currently).

If you're ever looking to add more tests, here are a few ways to get started:

  • run yarn test:unit and look at the coverage chart. You'll see all the uncovered lines. This is a good place to start.
  • look at test/scripts to see which scripts are tested. We can always use more tests there.
  • look at test/e2e. We can always use more end-to-end tests.

Otherwise, talk to a current maintainer and ask which part of the codebase is lacking most when it comes to tests.



Our docs are hosted on Vercel. Vercel only shows logs in realtime, which means you need to have the logs open in one tab and reproduce your error in another tab. Since our logs are private to Coder the organization, you can only follow these steps if you're a Coder employee. Ask a maintainer for help if you need it.

Taking a real scenario, let's say you wanted to troubleshoot this docs change. Here is how you would do it:

  1. Go to
  2. Click "View Function Logs"
  3. In a separate tab, open the preview link from github-actions-bot
  4. Now look at the function logs and see if there are errors in the logs

See an opportunity to improve our docs? Make an edit.