tulipfarm docsInstall (self-host)
One-line install of the TulipFarm OCI stack on Linux, macOS, or Windows (WSL2).
The one-line installer stands up the full TulipFarm stack (the app image + a bundled
PostgreSQL 17 + pgvector) with Docker or Podman, generates all secrets for you, and
prints the setup-wizard URL. It asks zero configuration questions.
Re-running the installer is the upgrade path: it preserves your .env (never
re-generates secrets) and your database, then pulls the pinned version and restarts.
Linux / macOS
curl -fsSL https://raw.githubusercontent.com/TulipFarm/tulipfarm/main/scripts/get.tulipfarm.sh | sudo bashWhat it does: detects a container engine (on Linux, auto-installs Podman if neither
Docker nor Podman is present), writes /opt/tulipfarm/{docker-compose.yml,.env},
generates ENCRYPTION_KEY / JWT_SECRET / WEBHOOK_SIGNING_SECRET / POSTGRES_PASSWORD,
brings the stack up, polls /health, and prints the URL to open.
On macOS with no container runtime, install Docker Desktop or Podman first. The no-VM native lane is not yet available.
Windows (WSL2)
From PowerShell:
irm https://raw.githubusercontent.com/TulipFarm/tulipfarm/main/scripts/install.ps1 | iexThis verifies WSL2 (and a Linux distro) is installed, then runs the Linux installer
inside WSL. If WSL is missing it prints the wsl --install steps to run first.
Finish setup
Open the printed URL. The app redirects to the setup wizard at /setup. Work through
the four steps:
- Admin account — enter an email and password. This is the only step that requires no authentication; the route locks once any admin exists.
- Business profile — give your instance a name and optional description.
- LLM provider — choose a provider (Anthropic, OpenAI, Azure OpenAI, or an OpenAI-compatible endpoint), enter your credentials, and pick a model. You can skip this step and configure it later in Settings → LLM Config.
- Soul backup — optionally set a git remote URL for soul backup and sync. Skippable.
After step 4, the wizard marks setup complete in soul.yaml and redirects to the main
workspace. The setup routes lock — no second visitor can claim the instance through the
wizard.
If the installer detected a public IP it prints a TLS warning: the instance is reachable over HTTP and the first visitor becomes admin — complete setup immediately and put a reverse proxy with TLS in front for production.
Once you are in, you run TulipFarm by talking to it — describe what you want and agents build it. See Building by chat to get started.
Overrides
Linux / macOS
All optional, set as environment variables before running:
| Variable | Default | Purpose |
|---|---|---|
TF_VERSION | latest | App image tag to run (ghcr.io/tulipfarm/tulipfarm:<tag>). |
TF_PORT | 8080 | Host port to publish. |
TF_INSTALL_DIR | /opt/tulipfarm | Install directory. |
TF_RUNTIME | auto | Force docker or podman. |
TF_BASE_URL / TF_REF | GitHub raw / main | Where compose + .env.example are fetched from. |
Example — pin a version on a custom port:
TF_VERSION=v0.1.0 TF_PORT=9000 \
bash -c "curl -fsSL https://raw.githubusercontent.com/TulipFarm/tulipfarm/main/scripts/get.tulipfarm.sh | sudo -E bash"Windows (WSL2)
The PowerShell script accepts named parameters. Download it first, then run with the parameters you need:
Invoke-WebRequest https://raw.githubusercontent.com/TulipFarm/tulipfarm/main/scripts/install.ps1 -OutFile install.ps1
.\install.ps1 -Version v0.1.0| Parameter | Default | Purpose |
|---|---|---|
-Version | latest | App image tag (TF_VERSION inside WSL). |
-BaseUrl | GitHub raw | Base URL for fetched files. |
-Ref | main | Git ref under -BaseUrl. |
To set TF_PORT, TF_INSTALL_DIR, or TF_RUNTIME, open your WSL terminal and run the bash installer directly with those env vars instead.