Windmill
Self-hostable platform that turns scripts into workflows, internal apps, APIs and AI agents.
Write code in 20+ languages and Windmill generates the form, API, schedule and flow around it.
Windmill is a code-first platform for workflows, internal apps and AI agents, built on a Rust job queue backed by Postgres. It suits engineering teams that want to replace ad hoc scripts, cron jobs and internal admin tools with one reviewed system.
What it is
Each script's typed arguments become an auto-generated UI, HTTP endpoint and CLI command. You chain scripts into flows, build apps on top, and trigger them from schedules, webhooks, HTTP routes or Kafka. AI sessions build scripts, flows and apps in a workspace fork with a diff to review before deploying.
Who it's for
It suits developer-led teams with someone who can run Postgres, Docker or Kubernetes. It does not suit non-technical users who want a hosted no-code tool with nothing to run, and it is not plain open source in every part.
Key features
Auto-generated UIs
Script parameters are parsed to create a form, an API endpoint and a CLI command automatically.
Code in many languages
Python, TypeScript, Go, Bash, SQL, GraphQL, PowerShell, Rust, C#, Java and Ansible run as steps.
Workflows as code
Define flows in the visual editor, as YAML, or as TypeScript with tasks and approval waits.
AI sessions with review
AI builds in a workspace fork with live preview and a diff, and protection rulesets force an engineer to deploy.
Git sync and CLI
Two-way Git sync, a wmill CLI and a VS Code extension keep workspaces and local files aligned.
Sandboxed workers
Jobs run under nsjail and PID namespace isolation, pulled by tag by worker groups.
Up and running in 3 steps
Our honest verdict
What's great
- One platform covers workflows, internal apps, cron jobs and AI agents, all as ordinary code you can review.
- Business teams can build with AI while protection rulesets make engineers approve what reaches production.
- Docker Compose gets a local instance running from three files, and Helm charts cover Kubernetes.
- Data features include managed Postgres data tables, DuckDB scripts and DuckLake lakehouse tables.
Where it falls short
- You must run and maintain Postgres plus API servers and workers; the README suggests 1 worker per 1 vCPU and 1-2 GB RAM.
- The licence is mixed: AGPLv3 and Apache 2.0 parts plus proprietary enterprise code, and the Community Edition images include non-open code.
- Selling it, serving it as a managed service, or wrapping it as a feature of your product needs a commercial licence.
- Audit logs, SAML and a distributed cache are Enterprise Edition features, and some features such as pipelines are marked alpha.
Pricing & license
Self-hosted Community Edition is free within its limits; windmill.dev/pricing shows a free open-source tier (max 10 users with SSO) and Enterprise from $120/mo, with per-builder and per-worker pricing. Windmill Labs also offers a cloud at app.windmill.dev.
Open remote AI-training roles
880 open remote AI-training roles, updated November 15, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
Is Windmill fully open source?
Not entirely. The LICENSE file says code is variously AGPLv3, Apache 2.0 and proprietary for enterprise features, and the Community Edition images include non-open code.
Can I use it inside my company for free?
Yes. The README says using Community Edition internally or via its APIs is free within the software's limits and quotas.
What does it need to run?
Postgres, plus API servers and workers; Docker Compose and Helm charts are provided.
