The core of Wardrowbe, a wardrobe manager with AI tagging and outfit suggestions, is open source under the MIT License. The backend and web app are on GitHub and you can self-host them with Docker Compose, with no item caps and no license key. Wardrowbe Cloud is a separate hosted edition with its own roadmap; how the two differ is under self-hosted or cloud below.
You need Docker, Docker Compose, at least 4GB of RAM, and an AI endpoint. The compose file pulls pre-built images from the GitHub Container Registry, so no build tooling is required on the host.
git clone https://github.com/Anyesh/wardrowbe.git
cd wardrowbe
cp .env.example .env
# Point AI_BASE_URL at Ollama or any OpenAI-compatible endpoint,
# then pull the published images and start the stack.
docker compose pull
docker compose up -d
docker compose exec backend alembic upgrade head
curl http://localhost:8000/api/v1/healthThe web app is then on http://localhost:3000 and the API docs on http://localhost:8000/docs.
| Runtime | Docker and Docker Compose |
|---|---|
| Memory | 4GB RAM minimum |
| Architectures | linux/amd64, linux/arm64 |
| AI endpoint | Ollama (free, local), OpenAI, or any OpenAI-compatible API |
| Database | PostgreSQL, included in the compose stack |
| License | MIT |
Your database and images sit on your hardware. The one thing that can leave your network is the image sent for AI analysis, and that is entirely determined by the endpoint you configure. Point AI_BASE_URL at a local Ollama instance and no photo ever leaves the machine. Point it at OpenAI and photos go to OpenAI. This is a single line in your environment file and you can change it whenever you like.
We think this matters more than it first appears, because a wardrobe app ends up holding a photographed inventory of everything you own. We wrote about that in what your wardrobe data reveals.
The self-hosted build is free and has the core of the app. You maintain the server, apply updates, and supply an AI endpoint, and the quality of tagging depends on which model you point it at.
The two editions started as one app and share the same core, but they now have their own roadmaps, so a feature in one is not always in the other. The cloud has, for example:
Ideas and core fixes move between the two. Contributions to the self-hosted edition stay free and open there and never go behind a paywall. If you want a cloud feature in the self-hosted edition, please open an issue or discussion on GitHub; if there's strong interest, we'll look into bringing it over. The self-hosted edition also has things the cloud does not, such as its eight languages. More in this discussion.
The cloud service is for people who would rather not run a server, and its subscriptions pay for continued work on the open source project. We compared the two honestly in cloud versus self-hosted, and put Wardrowbe next to the other self-hostable options in the best self-hosted wardrobe apps.
Yes. The core of Wardrowbe is open source under the MIT License and on GitHub: the backend, the web app, AI tagging and outfit suggestions. You can self-host it with Docker Compose, with no item caps. Wardrowbe Cloud adds virtual try-on and partner store suggestions, and the iOS app and the Android app are published separately; none of these are in the repository. OpenWardrobe is another open source option, though it is a manual cataloger with no AI or outfit generation.
Docker and Docker Compose, at least 4GB of RAM, and an AI endpoint. The AI endpoint can be Ollama running locally for free, OpenAI, or any OpenAI-compatible API. Published images are multi-arch for linux/amd64 and linux/arm64, so a Raspberry Pi or an ARM NAS works without building anything from source.
They share the same core: AI tagging, outfit suggestions, weather, households, notifications and wardrobe analytics, with no item caps and no license key. Past that core they are two editions with their own roadmaps, so features are not ported one to one. Wardrowbe Cloud has features that depend on hosted infrastructure or business agreements, such as virtual try-on (GPU servers), partner store suggestions (agreements with stores and affiliate networks) and Sign in with Apple and Google, plus features built for the cloud service first, such as receipt import, adding items from a text description and achievements. The self-hosted edition has community contributions of its own, and the GitHub README describes what it does today. The iOS app and the Android app are not open source, but both can connect to a self-hosted server. When you self-host, you run and update the server yourself and supply your own AI endpoint, which determines tagging quality.
Your photos and database stay on your hardware. Whether an image leaves your network depends on the AI endpoint you configure. Point it at a local Ollama instance and no image ever leaves the machine. Point it at OpenAI and images are sent to OpenAI for analysis. You choose this in the environment file and can change it at any time.
The MIT License, which permits commercial use, modification, distribution, and private use, and requires only that the copyright notice is preserved.
Issues and pull requests are welcome. The most recent release added eight-language support to the self-hosted build, which started as a contributor pull request. Star the repository if you find it useful, it genuinely helps other people find it.