> ## Documentation Index
> Fetch the complete documentation index at: https://docs.nuphos.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Self-host

> Run Nuphos on your own infrastructure with Docker Compose, Kubernetes (Helm), or Railway.

Nuphos is open source, so you can run the whole stack yourself: MongoDB,
S3-compatible storage, the backend, and an agent runtime. The deployment guides
live next to the code in [nuphos/nuphos](https://github.com/nuphos/nuphos) and
are kept up to date there, so this page only helps you pick one.

| Option | Use it for | Guide |
| - | - | - |
| **Docker Compose** | Trying Nuphos on one machine with a single `docker compose up` | [`deploy/compose`](https://github.com/nuphos/nuphos/blob/main/deploy/compose/README.md) |
| **Kubernetes (Helm)** | Running on a cluster you already operate | [`deploy/helm`](https://github.com/nuphos/nuphos/blob/main/deploy/helm/README.md) |
| **Railway** | A managed deployment without your own servers | [`deploy/railway`](https://github.com/nuphos/nuphos/blob/main/deploy/railway/README.md) |

## Before you plan around it

* **Email sign-in needs an email provider.** The backend sends sign-in codes
  by email, so the Helm and Railway guides ask for a Resend API key and a
  verified sender.
* **The guides set up email sign-in only.** Google and Apple sign-in need
  their own OAuth configuration on your backend.
* **The runtime signs in to Claude itself**, once, from its own console. You
  use your own account; there is no key to configure in Nuphos.
* **Point the apps at your backend from the sign-in screen.** Desktop and iOS
  have a **Self-hosted? Set API endpoint** link there. The Android app supports
  Nuphos Cloud only. For the [CLI](https://docs.nuphos.ai/cli.md), set `NUPHOS_API_URL`.
* **Cloud accounts are bound to your deployment.** AWS, GCP and the other cloud
  providers must trust *your* backend's identity, not nuphos.ai's. See the
  [IAM trust model](https://docs.nuphos.ai/security/iam-trust.md).
* **Not included:** managed runtime provisioning (**New computer**), managed
  backups, and the operations tooling of the hosted service.

## Operate and upgrade

Record the backend, Desktop/CLI and runtime versions before upgrading. Use the
release notes and the selected deployment guide for compatible image tags and
configuration; test the client sign-in and a bounded runtime task after rollout.

Keep persistent MongoDB data, object storage and runtime files in the volumes or
services described by that guide. Back up data and required configuration before
an upgrade, and plan restoration before assuming an older image can read a newer
database schema. A restart is not a backup, and downgrading an image alone may
not undo a data migration.

For cloud federation, configure your backend's issuer/WIF and use the values its
binding dialogs provide. OAuth callbacks, public backend URL and outbound access
must match your deployment; hosted-service trust values are not a substitute.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.