sc PR environment

One someones.computer deployment per pull request, a persistent staging deployment that tracks your default branch, a production deployment that tracks a prod tag, and one deployment per other branch — all from a single job, no per-trigger config. Wraps the sc CLI.

Quick start

You need an application already created on the platform, and an API token from that app's organization (Settings → API tokens) saved as a repo secret (SC_TOKEN below).

# .github/workflows/deploy.yml
name: Deploy

on:
  pull_request:
    types: [opened, synchronize, reopened, closed]
  push:
    branches: ['**']
    tags: [prod]

concurrency:
  group: sc-${{ github.event.pull_request.number || github.ref_name }}
  cancel-in-progress: ${{ !startsWith(github.ref, 'refs/tags/') }}

jobs:
  deploy:
    runs-on: ubuntu-latest
    permissions:
      pull-requests: write
    steps:
      - name: Deploy to someones.computer
        uses: https://git.grey.ooo/Grey.ooo/someones.computer_pr-environment-action@main
        with:
          app: my-app
          token: ${{ secrets.SC_TOKEN }}

That's the whole setup. One job, no mode, no deployment — both are inferred from what triggered the run:

Trigger	Deployment	Mode
PR opened / synchronized / reopened	pr-<number>	up
PR closed	pr-<number>	down
Push to the default branch	staging	up
Push of the prod tag	production	up
Push to any other branch	branch name, sanitized	up
Deletion of any branch/tag above	matching deployment	down

Once a deploy finishes, it also posts (and keeps updated) a PR comment with the deployment's live endpoint URLs — that's what pull-requests: write is for.

uses: needs the full URL, not owner/repo@ref — this Forgejo instance resolves a bare owner/repo against github.com, not itself. Pin to a commit SHA instead of @main for reproducible CI.

Inputs

Only app and token are required — everything else has a sensible default.

Input	Default	What
app	—	Application slug (required)
token	—	API token, pass as a secret (required)
mode	inferred	up or down — override the trigger-based inference
deployment	inferred	Deployment name — override the trigger-based inference
staging-branch	repo default branch	Branch that maps to the staging deployment
production-tag	prod	Tag that maps to the production deployment
comment	true	Post the deployment report on the PR
github-token	github.token	Token for the PR comment (needs pull-requests: write)
endpoint	production	sc --endpoint (platform API base URL)
staging	false	sc --staging shorthand — a platform endpoint, unrelated to the staging deployment name above
org	—	Disambiguates a token whose user belongs to several orgs
checkout	true	Check out the triggering commit before deploying
channel	stable	sc release channel to install
extra-args	—	Extra arguments appended to sc deploy

Outputs

Output	What
mode	The resolved mode (up or down)
deployment	The resolved deployment name
endpoints	Raw JSON of the deployment's live endpoints from the platform API

More

• Exact trigger-resolution rules, the deployment report, and how to split this into separate jobs instead of one: docs/reference.md.
• Who's actually using this, and how each app is wired up: docs/registry.md.
• Why PR environments exist, and the platform side of this action: docs/pr-environments.md in the platform repo.
• This is a separate public repo (not inside the private platform repo) because Forgejo clones a cross-repo uses: action anonymously — a private source means every consumer's CI gets a 403.