windmill-labs/windmill

What Windmill does

Windmill is a code-first orchestration platform for scripts, workflows and internal apps. A script is an ordinary function in Python, TypeScript, Go, Bash, SQL, Rust or another supported runtime; its typed arguments generate a form, an HTTP endpoint and a CLI command, and scripts can be chained into flows or used as steps in React/Svelte apps and AI sessions. The backend is written in Rust and workers pull jobs from a Postgres queue. The project positions itself as an open-source alternative to n8n, Retool and Lovable, with a lighter-weight scope than Temporal.

The same instance is meant for two audiences: engineers write code locally or via Git sync, while business teams build through browser-based AI sessions. Review steps, protection rulesets, dev workspaces and worker groups are there to keep deployment under engineer control.

Self-host with Docker Compose

Windmill provides a three-file Docker Compose setup: docker-compose.yml, Caddyfile and .env. From a directory where you want to keep those files:

curl https://raw.githubusercontent.com/windmill-labs/windmill/main/docker-compose.yml -o docker-compose.yml
curl https://raw.githubusercontent.com/windmill-labs/windmill/main/Caddyfile -o Caddyfile
curl https://raw.githubusercontent.com/windmill-labs/windmill/main/.env -o .env
docker compose up -d

Then open http://localhost and sign in with the default credentials admin@windmill.dev / changeme. Change them before exposing the instance. To use a managed Postgres such as RDS, Cloud SQL or Neon, set DATABASE_URL in .env and set database replicas to 0. Kubernetes deployments are available through Helm charts in the windmill-labs/windmill-helm-charts repository.

For a hosted trial instead of self-hosting, the project points to its app instance; the self-host documentation and Docker Compose section cover the detailed setup.

Main concepts and tooling

The workflow is script → generated UI/API → flow → app. You can author scripts in the web editor or sync a workspace to Git through the CLI, the VS Code extension or the MCP server, and local code can run against instance resources by configuring environment variables for the wmill client library. Data tables provide managed Postgres tables without exposing a connection string; DuckDB and DuckLake support querying files and lakehouse tables, and existing dbt projects can run unmodified as scripts.

Constraints

• The default deployment needs Docker and a Postgres database; one worker per vCPU with 1–2 GB RAM is the README's rough sizing rule.
• Runtime coverage differs by image: Deno is limited to windmill-full images, and PHP is available only there; the default JavaScript runtime is Bun.
• License terms are split: the Community Edition is free for internal use within its limits, but selling or wrapping it as a managed service needs a commercial license. The binary built without the enterprise feature is AGPLv3; API clients and the OpenAPI spec are Apache 2.0.
• The project states it is not seeking outside contributions at this time, though bug reports and small verified fixes remain welcome.
• The "fastest self-hostable workflow engine" claim comes from the project's own benchmark against Airflow, Prefect and Temporal; the methodology is on its benchmark page rather than in the README.

HotRepo's take

Windmill has accumulated roughly 18.4k stars, and its most visible community moment was a 2023 Hacker News thread that drew 428 points and 109 comments. That discussion is useful context rather than a verdict: one commenter praised the pace of fixes and the project's Discord community, another suggested the documentation should lean more on code-oriented material and IDE/Git workflows than on the web editor, and a third questioned whether a single tool spanning many jobs conflicts with Unix-style composability. The repository itself has one recent recorded ranking on the Rust daily list (18th, 31 stars gained on 2026-11-06), with no all-language daily rank recorded. Readers weighing adoption should evaluate the license terms and the AGPL/commercial split carefully, since those affect how the platform can be offered to others.