- Home
- VPS Hosting
- LibreChat VPS
LibreChat pre-installed · the app is free
LibreChat Hosting on Your Own VPS
Devoster's LibreChat hosting is a KVM VPS with NVMe storage in Utah, USA with LibreChat pre-installed: one chat interface for OpenAI, Anthropic, Google, OpenRouter, Mistral and more, where each user enters their own API keys, stored encrypted. The app is free; plans start at $4.49/mo, with Starter (2 vCPU, 4 GB, $7.99/mo) recommended.
- LibreChat 0.8.8
- Many providers, your own keys
- Agents, presets and search
- Registration off by default
- KVM · NVMe · full root · Utah, USA
Prices verified
Recommended plans
- Recommended
Starter
$7.99/mo- 2 vCPU
- 4 GB RAM
- 75 GB NVMe
- 1 IPv4
Micro
$4.49/mo- 1 vCPU
- 2 GB RAM
- 50 GB NVMe
- 1 IPv4
Basic
$10.99/mo- 3 vCPU
- 6 GB RAM
- 100 GB NVMe
- 1 IPv4
What you get with one-click LibreChat
LibreChat is an open-source chat interface for AI models from many providers. You switch between OpenAI, Anthropic, Google and others in one place, with your conversations stored on your server. Choose it with a plan, pay, and we build the VPS (KVM, NVMe, Ubuntu 24.04 LTS) and install it in about five minutes. You get:
- LibreChat 0.8.8 (MIT) with MongoDB for conversations and Meilisearch for searching them, neither reachable from the internet.
- An HTTPS address like
yourname-x7k2.devoster.appwith a Let's Encrypt certificate. - Your admin account already created, logging in with your Devoster account email and a generated password shown in the dashboard.
- Registration off from the very first start, so the public address never shows a sign-up form.
- Dashboard controls to open the app, reset the admin password, restart, stop or start it, read its logs, run an update and connect your own domain.
The app is free; you pay only the VPS plan, with 10% off yearly. Model usage is billed by your providers, per token, to your own keys. The server comes with full root, free nightly backups kept 3 days plus one snapshot, a dedicated IPv4, 1 Gbps with unlimited bandwidth and DDoS protection. Already have a Devoster VPS? Install LibreChat on it from VPS Control, after a quick check of the OS, ports 80 and 443 and free disk, or reinstall the server with it; both are free. Other apps that install this way are on the One-Click Apps page.
Bring your own API keys: which providers work
LibreChat doesn't include a model or any credit. Each user picks a provider in the model selector, chooses Set API key and pastes their own key. LibreChat stores it encrypted with a key generated on your server. Ready to use on this install:
- Built in: OpenAI, Anthropic and Google.
- Added as custom endpoints: OpenRouter, Mistral, Groq, DeepSeek and xAI.
Because keys are per user, everyone pays for what they use on their own provider account, and nothing is billed to the server. To add another OpenAI-compatible provider, or give one of the custom endpoints a shared key, edit /opt/devoster/data/librechat/librechat.yaml and restart the app from the dashboard. That file is yours after the first install and updates don't overwrite it.
Devoster VPS plans are CPU-only, with no GPU, so this install doesn't run local models. A provider you pay per token is what makes answers fast.
Agents, presets, search and sharing
Beyond plain chat, this install comes with:
- Agents: a model with your instructions and tools, saved and reused.
- Presets that store a model with its settings and system prompt, so a team reuses the same setup.
- Conversation search across your history, powered by Meilisearch.
- Share links to show a conversation to someone else.
Not set up: file search and chatting with your documents (RAG). In LibreChat that needs a separate service with a server-side embeddings key, which this install leaves out. If you need a document chat, it is a project to add yourself with root access, following LibreChat's documentation.
LibreChat or Open WebUI?
We offer both, and they suit different teams:
- LibreChat is a front end for many providers' own APIs. Each user brings and stores their own keys, and you get OpenAI, Anthropic and Google natively plus several others side by side.
- Open WebUI grew up around local models through Ollama, with OpenAI-compatible connections that the admin sets up for everyone. It feels closer to a single shared workspace.
Pick LibreChat when people want to use their own keys across several providers. Pick Open WebUI when an admin should set up the models once for the whole team.
Adding users, plans and email
Adding people. Registration is off, so new users are created over SSH with LibreChat's own script: from /opt/devoster/app, run docker compose exec api npm run create-user and answer its prompts. Each person then logs in with their email and enters their own provider keys.
No email. Outbound mail ports are blocked by default on Devoster VPS plans and LibreChat has no mail server configured, so there are no password-reset or verification emails. Reset the admin password from the Devoster dashboard.
| Your use | Plan | Why |
|---|---|---|
| One person trying it out | Micro (1 vCPU / 2 GB) | The smallest plan it is offered on. LibreChat, MongoDB and Meilisearch fit, tightly. |
| Daily use, a small team | Starter (2 vCPU / 4 GB) | More memory for the database cache and search index. |
| A larger team with long histories | Basic (3 vCPU / 6 GB) | A third core and room for a bigger search index. |
The models run at your providers, so the VPS only serves the interface and stores history. You can upgrade from the panel later without reinstalling.
Updates, licence and what stays your responsibility
LibreChat moves quickly. We test new versions and apply them in a maintenance hour you choose, or when you press Update now. A snapshot of the VPS is taken first, and if LibreChat doesn't come back healthy the server is rolled back automatically. Conversations, users, stored keys and your librechat.yaml carry over. Auto-update can be switched off.
LibreChat is MIT-licensed, and Devoster isn't affiliated with it or with any of the model providers named on this page.
The server is unmanaged: you have full root, and its security, the accounts you create and the data people send to model providers are your responsibility. Self-hosting keeps conversations on your VPS, but every message still goes to the provider that answers it, under that provider's terms. Nightly backups (3 days) cover the whole VPS.
Frequently Asked Questions
Related pages
- Open WebUI VPSOpen WebUI hosting on a KVM VPS in Utah, USA: a private ChatGPT-style chat for your own API keys, pre-installed with HTTPS. From $4.49/mo, Starter $7.99/mo.
- n8n VPSn8n VPS hosting with n8n and PostgreSQL pre-installed on KVM + NVMe in Utah, USA. HTTPS in minutes, n8n updates done for you. Starter $7.99/mo, full root.
- Activepieces VPSActivepieces hosting on a KVM + NVMe VPS in Utah, USA: Activepieces, PostgreSQL and Redis pre-installed with HTTPS. Starter $7.99/mo, no per-task fees.