Skip to main content

Self-Hosted Guide

Background​

The self-hosted version of Wren AI lets you run the platform on‑premises or in your own cloud. Feature availability varies by plan - contact us for details.

Hardware Requirements​

Minimum production system requirements:

  • vCPU: 8+
  • Memory: 32 GB+
  • Disk: 128 GB+ (used for system logs, the application database, and dashboard cache; grows with usage)

Network:

  • Ensure HTTP (port 8080) or HTTPS (port 443) are accessible as required for your environment (public or private access)

Supported OS​

  • CentOS Stream ≥ 9
  • Red Hat Enterprise Linux (RHEL) ≥ 9

We ship CentOS as the default VM image OS. If you prefer to install on RHEL from scratch, we provide an installation script - just run it to complete setup.

Installation Guide​

1. Launch VM with the image​

We provide you with a ready-to-use virtual machine image. For AWS users, this comes as an Amazon Machine Image (AMI).

You simply launch a new virtual machine using this image. The VM already comes pre-installed with Wren AI.

2. Start Using Wren AI​

Once the VM is running, you can connect to it directly from your browser. The first thing you’ll see is the license setup page. Just paste in the license key provided during your purchase, and you’re ready to start using Wren AI immediately.

FAQs​

Q1. Why is it a VM image and not a container image?

Wren AI itself is a microservices-based architecture where each service runs in its own container. Instead of asking you to set up and orchestrate multiple containers manually, we package the entire stack into a single VM image. This way, all services are already bundled, configured, and optimized, so you can get started faster.

Q2. Do I need Kubernetes or Docker knowledge to run Wren AI?

No. All the underlying container orchestration is handled inside the VM image. You only need to launch the VM and connect via your browser.

Q3. How do I obtain the license key?

The license key is issued to you during the purchase process. You just need to paste it into the UI on first login.

Q4. Can the deployment scale if my workload grows?

Yes. Wren AI runs on a microservices architecture, and inside the VM all services are containerized. This means you can scale those containers to handle larger workloads directly within the VM. Since we leverage k3s (a lightweight Kubernetes distribution) as the underlying orchestration layer, the system is compatible with Kubernetes architectures. If your usage grows beyond what a single VM can support, we can work with you on other Kubernetes-based deployment options to scale horizontally across multiple nodes.

Q5. Which LLMs are supported?

Supported providers: OpenAI and Azure OpenAI; Google Gemini; and Anthropic Claude via AWS Bedrock.

Recommended primary models for core analysis and quality:

  • Gemini 2.5 Pro
  • Claude 4 Sonnet
  • OpenAI GPT‑4 class (e.g., GPT‑4o)

Use fast/cost‑optimized variants (e.g., Gemini 2.5 Flash, 2.0 Flash Lite; Claude 3.5 Haiku) selectively for non‑critical tasks.