MLflow Hosting Live In 5 Minutes

MLflow hosting for teams that need a private place to track experiments and compare runs. Deploy MLflow 3.15.2 as a self-hosted machine learning experiment tracking server: xCloud runs the pinned image behind an authenticated HTTPS gateway, keeps metadata and artifacts on one persistent volume, and generates the credentials you sign in with, so nothing in the interface is reachable anonymously.

Deploy your MLflow

No terminal
No technical skills needed
No lock-in, cancel anytime

What is MLflow?

Self-hosted MLflow is a machine learning experiment tracking server. Its README describes MLflow as an open source AI engineering platform for agents, LLMs and ML models, and lists experiment tracking of parameters, metrics and results, model evaluation, a model registry and deployment tools. Teams run their own server instead of a hosted tracker such as Weights and Biases so that runs and artifacts stay on infrastructure they control.

On xCloud the template runs the ghcr.io/mlflow/mlflow v3.15.2 image behind a Caddy gateway, both pinned by immutable digest. MLflow stores metadata in SQLite and serves artifacts itself, all on the mlflow_data volume, and only the gateway is published, over HTTPS. Caddy enforces Basic Authentication on every request using the administrator username you choose and a generated password. The setup is a compact single-node baseline.

From Sign-Up to a Live MLflow Workspace In Just 3 Easy Steps

Self-hosting MLflow normally means a container, an authentication layer in front of it, persistent storage for metadata and artifacts, and a domain and certificate before a single run is logged. This is the same result without assembling any of it.

Step 01
Sign Up
Create your xCloud account and pick MLflow from the OneClick catalog. Enter a site title and domain, confirm an administrator username, and let xCloud generate a strong password.
Step 02
We Deploy It For You
One click. We bring up the official MLflow 3.15.2 image, pinned by immutable digest, behind a Caddy gateway that protects the entire UI, tracking API and artifact API with your generated credentials, mount the mlflow_data volume for SQLite metadata and served artifacts, and publish it over HTTPS on your domain.
Step 03
Sign In And Log Your First Run
Open your HTTPS URL and sign in with the generated administrator credentials shown in your xCloud credential panel. Create an experiment, start a run, log a parameter and a metric, upload a small artifact, and read it all back to confirm the workspace works end to end.

Everything MLflow Does, On Your Own Server

MLflow runs in full here — experiment tracking, run comparison and artifact storage, on infrastructure you control.

Experiment And Run Tracking
Create experiments, start runs and browse the full run history from the familiar MLflow interface.
Metrics, Parameters And Tags
Log and compare metrics, parameters and tags across runs to see what actually changed between them.
Artifact Storage And Retrieval
Upload and read back artifacts — files, models, plots — served directly by MLflow from its own volume.
Authenticated Access, Front To Back
A Caddy gateway protects the entire UI, tracking API and artifact API with your generated credentials — nothing is reachable unauthenticated.
MLflow Stays Off The Public Network
The MLflow container itself has no path to the internet — only the authenticating gateway is published.
Persistent Metadata And Artifacts
SQLite metadata and served artifacts share one persistent volume, so a run and everything it produced stay together.

Everything You Need To Run Your Experiment Workspace 24/7

Managed hosting for MLflow, so the parts that keep it online are not your problem.

MLflow 3.15.2 deployed instantly
Official pinned image digest
Administrator credentials generated for you
Authenticated gateway in front of the full API
Free SSL certificate
Server backups and snapshots
Server security updates included
30+ server locations
Unused balance refunded

What You Can Do With MLflow

Compare training runs side by side
Log parameters and metrics from each run to your own server and open the MLflow interface to see which change moved the result, with every run kept in the SQLite database on your Cloud VPS.
Keep model files next to their runs
Upload plots, models and other artifacts through MLflow, which serves them from the same volume as the metadata, so a run and the files it produced are retrieved and backed up together.
Give a small team one tracking URL
Share a single authenticated HTTPS address with colleagues so everyone logs to the same server, rather than keeping separate local MLflow folders on laptops.
Trace LLM application experiments
Point an LLM application's tracing at the server to inspect calls, per MLflow's tracing features, while keeping prompts and outputs off a third-party service, at modest single-node volumes.

MLflow vs Alternatives

Weights and Biases: A hosted experiment tracking service. MLflow's README describes an open source platform you run yourself, so with this template runs, metrics and artifacts sit on your own Cloud VPS instead of a vendor's account.
Langfuse: Langfuse is a self-hosted LLM observability tool with its own xCloud template. MLflow's README spans experiment tracking, a model registry and LLM tracing, so choose by whether you also train models.
Kubeflow: Kubeflow is a Kubernetes-based ML platform. This template is far smaller: a single-node MLflow server and gateway with SQLite storage, with no cluster to operate.
ClearML: ClearML is another self-hosted experiment tracker people compare. The MLflow template here adds mandatory Basic Authentication through Caddy and keeps everything on one backed-up volume.

Simple Pricing. Everything Included.

Dedicated Cloud VPS sizes, pre-configured for MLflow and its authenticated gateway. Pick the one that fits, and resize any time.

Cloud VPS 16 GB

Cloud VPS 24 GB

More concurrent experiments and a growing artifact store.

$84.99per month

24 GB RAM
8 vCPU cores
300 GB NVMe SSD
30 TB
Unlimited

Get started

Every size is a dedicated Cloud VPS we provision, patch, monitor and restore. Billed monthly with no lock-in — delete the server and the unused balance comes back to your account. The 6 GB rung sold elsewhere on this site is deliberately absent: this app's own template minimum is 3 GB, which does not leave comfortable headroom on it once the operating system is counted. MLflow itself is free under the Apache License 2.0.

Docker hosting

Already Have a Server? Bring Your Own

Connect any VPS — DigitalOcean, Vultr, Linode, AWS, Hetzner or bare metal — and deploy MLflow on it through the same panel, for a per-server management fee instead of a Cloud VPS rental. Size it above MLflow's own 3 GB minimum.

Free

Start your journey without cost.

$0forever

1 server
10 sites

Start free

Starter

Professional

The same platform, discounted at 6–10 servers.

$4per server / month

6–10 servers
Unlimited sites
All features included
Bring any VPS provider

Get started

Agency

The same platform, discounted at 11+ servers.

$3per server / month

11+ servers
Unlimited sites
All features included
Bring any VPS provider

Get started

Priced per connected server, so your rate adjusts automatically as you add or remove servers. You pay your own provider for the server itself.

Self-managed hosting

Ready To Run Your Own MLflow Tracking Server?

MLflow on xCloud runs on a dedicated Cloud VPS that we provision, patch, monitor and restore. Deploy faster. Own the server. Stay in control.

Fast Provisioning
Deploy faster using an xCloud Managed Server built with optimized stacks and high-speed SSDs.
Enterprise-grade security
Protect your infrastructure with a free SSL certificate and more advanced security features at no additional cost.
Intuitive Control Panel
From Server creation to application deployment, everything can be managed from our intuitive control panel.

Hear what people are saying about us

See why agencies, developers, startups, and growing businesses choose xCloud to power their projects.

xCloud works really well, I am surprised by its performance being this the release version, I have tried many paid platforms such as runcloud, cloudpages or serveravatar, as well as free platforms such as CyberPanel, CloudPanel or Hestia Control Panel and direct competitors such as SpinupWP and FlyWP. Of all those that I have tried, the platform that works best for WordPress is SpinupWP and I believe that xCloud can reach and surpass that development, its roadmap is promising. Congratulations to the xCloud team, I hope they become the leading platform in the segment, it has all the potential to achieve it!

Brian Geovanny

I've been with runcloud for the last 3 years. However, after trying xcloud for the last month, I'm switching over. They are on a roll with their setup.

Kevin Pineda

I'm moving my servers to xCloud, a cloud control panel by M Asif Rahman and the WPDeveloper team. Their roadmap looks interesting and no, there are no team mates nor API limits with xCloud! #xCloud #wordpress #hosting

Leo Koo

How Cloud Nine Web Elevated Client Experience While Streamlining Hosting Operations using xCloud

Ed Ellingham

xCloud is like having an expert server team on call — without the big agency costs. It's given me back my time and given my clients peace of mind. Highly recommended for freelancers and small teams who want reliable performance and security.

Dumitru Butucel

xCloud isn't just a solution. It's like one of that friend who got your back when you screw things up pretty bad.

Waqar Ahmed

Trust me when I say, once people explore the potential of xCloud they gonna switch from other panels, just keep delivering the best tech, lots of love for the team 😍😍

Rahul Singh

So far my experience has been smooth like silk. I didn't imagine that configuring and managing a VPS would be so easy. As of today, I have deployed 8 sites. The support is good, and they treat users with respect, even if they ask stupid questions (myself included). For the future, I hope for greater flexibility in scheduled backups and more comprehensive and consistent documentation. I'm sure that within a few months, xCloud will become the leader in its segment. This community is amazing, and the Staff here is super helpful.

Massimo Villa

Came from cloudways then to runcloud then to xcloud. Looks like I'm staying here.

Gerson El

I've recently taken all of my Vultr-hosted sites and gotten rid of cPanel for them to substitute it with xCloud. I have to say, I'm really impressed with xCloud so far! The interface is definitely the standout feature for me.

Tony Lewis

I came from Cloudways and won't go back. Highly impressed with xCloud. Glad I gave them a chance.

Jared McDowell

I like xCloud better than RunCloud, but one thing I'll point out is that some years ago RunCloud was hungry for business, the support was outstanding, and the community was active and helped each other. Maybe it was sold, there was a change in management, or they stopped being hungry, but the quality of the RunCloud offering declined and it hasn't been moving forward quickly. My point is that currently the xCloud team has the edge and it is theirs to keep or lose

David McCan

How Kevin Heinrichs Cut Hosting Costs by 84% While Running WooCommerce & Node.js Sites Efficiently

Kevin Heinrichs

I had 150 sites on cloudways and am in the process of moving them to xcloud. Much faster and better priced.

Dan Spinoza

Dear xCloud Team, I have been using xCloud for some time now and am genuinely impressed with its capabilities and the flexibility it offers by allowing users to bring their own VPS for management. It's a fantastic service, and my experience so far has been very positive. Successful migrated 5 sites on a single server. Xcloud migration tool worked great. There were few hiccups with security plugin like MalCare but disabling the plugin prior to migration worked just fine.

Rahil Maknojia

Came from runcloud. Tried xcloud and never came back.

Karl Fanega

How Advanced Web Strategies Saved Over $2,000 & Recovered 56 Hours per Month Managing 70+ Client Websites

Gary Moody

xCloud gave me what no other hosting platform could: the flexibility of a VPS with the convenience of managed WordPress. As a solo founder running an entire business on one server, that combination is an important unfair advantage.

Jorge de los Reyes Martínez

I use Server Avatar and xCloud and I'll sooner use xCloud than Server Avatar personally.

Jay Pagano

How Newman Web Solutions Ensured Perfect Uptime and Seamless Scaling For 120+ WordPress Sites with xCloud

Lori Newman

I too have moved over to xCloud. It's pretty sweet. One click n8n and uptime Kuma apps

John Aperture

I went from Cloudways to xCloud. No intention on switching in the foreseeable future.

Blaine Moore

Frequently Asked Questions

Feel free to contact us with any questions or concerns you may have.

How much RAM does MLflow need?

The template manifest asks for 3072 MB of RAM, 2 CPU cores and 20 GB of disk as a floor. A 6 GB server would leave only about 3 GB above the app's own floor once the OS is counted, so this page sells the 16 GB tier as the entry rung rather than calling that comfortable.

Is MLflow free and open source?

Yes. MLflow is Apache-2.0 licensed, per the LICENSE.txt at the v3.15.2 tag, and its README calls it an open source AI engineering platform. There is no licence fee; xCloud charges for the dedicated Cloud VPS and deployment around it, and the experiments stay yours.

How do I self-host an MLflow tracking server with Docker?

This is MLflow Docker hosting prepared for you: the template runs the official ghcr.io/mlflow/mlflow v3.15.2 image and a Caddy gateway, both pinned by immutable digest, with the mlflow_data volume and HTTPS on your domain. You point your code's tracking URI at the site and log in.

What is the MLflow backend store?

It is where run metadata lives. Here the template starts MLflow with a SQLite file at /mlflow-data/mlflow.db for experiments, runs, metrics, parameters, tags and the registry. That suits a compact single-node workspace; no PostgreSQL driver is in the image and none is added.

Where does MLflow store artifacts?

MLflow serves them itself and writes them under /mlflow-data/artifacts, on the same mlflow_data volume as the SQLite database. Runs therefore stay together with the files they produced, and one volume backup captures both. External object storage is not configured by this template.

Does MLflow support authentication?

Not through MLflow's own auth app here: the upstream image omits a package it needs, so a Caddy gateway enforces HTTP Basic Authentication in front of the UI, tracking API and artifact API with your generated credentials. MLflow also restricts allowed hosts and CORS origins to your domain.

Can I trace LLM applications with MLflow?

MLflow's README lists LLM tracing, evaluation, prompt management and an AI Gateway among its features. This template runs the server in a compact single-node setup with SQLite, so it suits modest tracing volumes; it is not configured for high-concurrency production traffic.

How do I upgrade MLflow?

Nothing upgrades it for you: the template pins MLflow v3.15.2 by immutable digest. Read the migration and release notes, take a consistent backup of mlflow_data, then verify sign-in and run read-back after raising the pinned image, or adopt a newer template version.

Can this template handle high-concurrency or team-scale workloads?

Not as configured. SQLite metadata and local served artifacts fit a compact single-node workspace. The image has no database or object-storage drivers and this template adds none, so larger multi-worker deployments need an upstream-supported external database and artifact store.

Is this MLflow cloud hosting?

It is your own MLflow on a dedicated Cloud VPS that xCloud provisions, not a shared tracking service. The server, the mlflow_data volume and the administrator credentials belong to your install, and only the authenticating gateway is published over HTTPS.

What should I back up?

A backup is the mlflow_data volume plus your Environment values, taken at the same point in time. The volume holds both the SQLite database and every served artifact, so restoring one without the other, or without the credentials, leaves runs pointing at missing files.

Backups in xCloud

Can I cancel anytime?

Yes. There is no lock-in — cancel any time. Hosting is billed monthly and charged for what you actually use, so when you delete the server the unused balance is refunded to your account, less up to 10% in payment-processing fees.