Host your AI agent on a VPS in France

AI VPS to run your agents continuously: root access, dedicated resources and 24/7 availability.

A VPS to host AI agents, not a GPU server: model inference stays with your provider, through your own API keys.

11,45 €6,87 €excl VAT/month

Order  

* You can increase your VPS resources at any time.

2 vCores
8 GB RAM
50 GB SSD NVMe
1 GBp/s bandwidth
5 minutes delivery

Checkout in 5 minutes!

Benefit from the validation of your VPS in less than 5 minutes* at Firstheberg! The validation of your order is automatic and your VPS is delivered within half an hour.

*Automatic delivery for customers residents of France, Belgium, Swiss and Luxembourg and subject to condition that the customer account shows no irregularity.

Features included with our VPS offer

Déploiement rapide Fast deployment
Deploy your VPS in less than 5 minutes with Windows or more than 10 Linux distributions.
Débits garantis Guaranteed speeds
Benefit from a guaranteed speed on the entire VPS offer ranging from 200 Mbps to 1 Gbps (10 Gbps optional) and a "burst" connection of 1 Gbps in WAN and LAN.
Reverse DNSReverse DNS
Change the reverse of your VPS directly from your customer area, thanks to our simplified tool for managing your DNS reverses.
GTR 4H GTR 4H
We commit to restoring your VPS within 4 working hours in the event of a hardware or network failure upon opening a support ticket on your client area.

Operating systems

A wide range of Windows & Linux OS available.
Choose and install the OS of your choice among the most popular on the market.

Firstheberg offers you the most popular operating systems for all of our VPS. You have the choice among several Linux distributions, such as Debian, CentOS Stream or Ubuntu, as well as Windows Server.

debian Debian
debian13 Debian 13
debian12 Debian 12.9
debian11 Debian 11.9
ubuntu Ubuntu
ubuntu2504 Ubuntu 25.04
ubuntu2404 Ubuntu 24.04 LTS
ubuntu2204 Ubuntu 22.04 LTS
rockylinux Rockylinux
rockylinux9 RockyLinux 9.5
rockylinux10 RockyLinux 10
almalinux Almalinux
almalinux9 AlmaLinux 9.5
suse Suse
suse15 Suse 15
cloudlinux Cloudlinux
cloudlinux96 Cloud Linux 9.6
fedora Fedora
fedora42 Fedora Cloud 42
fedora41 Fedora Cloud 41
centos Centos
centosstream9 CentOS Stream 9
centosstream10 CentOS Stream 10
windows Windows
win2022std Windows Server 2022 Standard
win2019std Windows Server 2019 Standard
Applications
WordPress WordPress
NextCloud NextCloud
ISPConfig ISPConfig
VestaCP VestaCP

What is a self-hosted AI agent?

A self-hosted AI agent is an autonomous program that runs on your own server and carries out tasks for you: monitoring a mailbox, calling APIs, generating reports, triggering scripts.

Unlike a chatbot conversation, which ends as soon as you close the window, an agent hosted on a server stays active permanently: it keeps its memory, holds its sessions open and runs its scheduled tasks even overnight.

That is exactly why a laptop is not enough: as soon as it goes to sleep or changes network, the agent stops. A VPS stays powered on, connected and reachable 24/7.

Why choose a VPS for your AI agent?

Full root access
An agent needs to install its dependencies, open ports and run system commands. You get unrestricted root SSH access on your machine.
KVM isolation
An agent that executes code deserves an environment separate from your personal machine and private data. KVM virtualisation fully isolates your VPS from other customers.
24/7 availability
Your scheduled tasks and recurring automations run continuously, on redundant infrastructure with 24/7 technical support included.
Data hosted in France
Your servers are hosted in our datacenters in France, which simplifies your GDPR compliance and reduces latency to your European users.

Self-hosted agent or AI subscription?

A consumer AI subscription charges a flat fee per user, imposes quotas and starts from a blank page with every conversation. An agent you host yourself works differently: you pay for your server and for your model calls based on actual usage.

Comparison between an AI subscription and a self-hosted agent on a VPS
Criterion Consumer AI subscription Self-hosted agent on a VPS
Usage limits Imposed quotas, access sometimes restricted mid-month No quota imposed by the host
Memory Starts from scratch every session Persistent memory, stored on your own server
Model choice Set by the vendor Free choice: an economical model by default, a premium model only when the task requires it
Cost structure Flat fee per user One VPS shared across your projects + your API keys billed on usage
Data Held by the vendor On your VPS, in our datacenters in France

The main saving lever is model choice: routing simple tasks to an economical model and reserving premium models for the cases that justify them substantially reduces your API bill. A single VPS can also run several agents and automations for your different projects.

Compatible with open source AI agents

Vous choisissez le runtime, vous l'installez et vous le configurez vous-même sur votre VPS. Les projets open source les plus utilisés fonctionnent sur une distribution Linux standard avec accès root.

OpenClaw
Autonomous agent
Runs tasks and scripts on your server.
Open Interpreter
Autonomous agent
Turns your natural-language instructions into running code.
Hermes Agent
Autonomous agent
Persistent agent with memory and scheduled tasks.
Paperclip
Orchestrator
Coordinates a team of agents: roles, budgets, recurring tasks.

An orchestrator relies on a PostgreSQL database and a scheduler that must run permanently: it needs a server that stays powered on, plus some memory for the database and the agents it triggers.

These projects are independent of Firstheberg: we provide the server, the connectivity and the infrastructure support, while you keep control of the software, its versions and its configuration.

FAQ - VPS for AI Agent

No, and that is not what it is built for. This VPS hosts the agent, meaning the program that takes your instructions, calls a model and carries out tasks. The model computation itself stays with your provider (Anthropic, OpenAI, Mistral and others) and is billed per use with your own API keys. An agent therefore needs no GPU and no accelerator card: it needs memory, a stable network and a server that stays on. Running an LLM locally requires a machine with a GPU, which our VPS range does not offer.
No. Your VPS is delivered with the operating system of your choice and full root access: you then install the AI agent runtime you prefer. The server is ready in under 5 minutes and you keep full control of your environment, your versions and your data.
Memory is the deciding factor: plan for at least 8 GB of RAM for a persistent agent that keeps a history, a vector index and several scheduled tasks. Since the model computation runs at your LLM provider over API, CPU power matters less than RAM and network stability. If you add a multi-agent orchestrator and its PostgreSQL database, plan for more memory and storage. You can adjust vCore, RAM and storage at any time from your customer area.
Yes. You have root access: you configure the model providers of your choice (Anthropic, OpenAI, Mistral, open source models, etc.) with your own API keys. No restriction is applied to the external services your agent contacts, subject to our terms of use.
It depends on your usage. You replace a flat fee with two variable costs: your VPS and your model calls billed on usage. The benefit is twofold: a single server can host several agents for your different projects, and you keep control over which model handles each task, which is the main saving lever. For occasional, light use a subscription may remain simpler; for an agent running continuously or across several projects, self-hosting quickly becomes worthwhile.

The other FirstHeberg VPS plans

Firstheberg est la marque commerciale de la société Techcréa Solutions, société créée en 2010, au capital social de 235 049 €. Firstheberg est le spécialiste de l’hébergement web, VPS Linux et Windows et serveur dédié pas cher et pour tous. N’hésitez pas à consulter nos offres d’hébergement mutualisé et noms de domaine offerts, VPS SSD Linux OpenVz, VPS SSD Linux et VPS SSD Windows Qemu, et Serveurs dédiés HDD et SSD.