sprout

Every pull request gets its own preview: an isolated database on one shared instance plus an optional live app, for teams that review running code instead of diffs. Self-hosted, no platform lock-in — everything is torn down when the PR closes.

  PR opened / push          PR closed
        │                        │
        ▼                        ▼
   sprout deploy ──► gateway ──► sprout teardown
        │                        │
        ├─ CREATE DATABASE       ├─ stop containers
        ├─ start app (+ Traefik) └─ DROP DATABASE
        └─ optional seed image
               │
               ▼
      preview URL for this PR

Adopt in three files

.sprout.yaml, a CI include, and the Dockerfile the app already ships — plus two CI variables from the operator. The published gateway image reports anonymous install telemetry by default (SPROUT_TELEMETRY=off stops it). Humans start at Getting started. Agents: paste the Onboarding prompt into a coding harness and it wires the repo. Live pitch: the marketing page.

Docs

• Public docs — guides and reference (source · llms.txt)
• Getting started — first preview in one sitting
• Adopting a repo — write .sprout.yaml, migrate at startup
• CI integration — GitHub / GitLab wiring, reset
• Previews — databases, seeding, services, mail
• Operator deploy — compose stack, env, Traefik
• CLI reference — every command and flag
• Troubleshooting — symptom → fix catalogue
• Telemetry — what is sent, what it costs, off switch
• Onboarding prompt — paste into a coding harness
• examples/adopting-repo/README.md — copy-paste starter
• CONTEXT.md — domain vocabulary
• e2e/README.md — acceptance harness (bun run test:e2e)

Develop

bun install
bun test
bun run typecheck   # turbo build + docs links

License

AGPL-3.0-only — see LICENSE.