A server for an AI agent

1,700 servers from 53 providers, from €4.49 a month. Prices checked 7 hours ago

A self-hosted AI agent such as OpenClaw or Hermes Agent is a small always-on program: it keeps your chat channels connected, runs tools and calls a model provider’s API to think. The model runs at the provider, so the server needs no GPU. OpenClaw’s documentation puts the absolute minimum at 1 vCPU and 1 GB of memory and recommends 2 GB or more for logs, media and several channels, so this list starts at 2 vCPU and 2 GB.

Lowest bill
€4.49(The lowest bill with every charge known. Rows priced lower leave out a charge their provider doesn’t publish.)
Typical hidden fees
+€6.24/mo(49% of these servers cost more than their price tag once the required fees are added (IPv4, setup, traffic, minimum term). On those, the extra is typically €6.24 a month. The list is always sorted by the full bill.)
Bill over
Bill horizon

1,700 servers match, sorted by bill per month.

25 servers, bills over 12 mo
Tick 2 to 4 servers to compare their bills side by sideTrafficLocationsTotalActions
phoenixNAP Bare Metal 2 × Intel Xeon Gold 5315Y · 2 × 4 TB NVMe · Singapore (d2.m1.xlarge) 16 cores · 32 threads · 512 GB RAM · 8 TB NVMe · 5 TB traffic +€26.73/TB Check due
16 cores 32threads 512 GB NVMe 8 TB 5 TB +€26.73/TB €719.98(lowest complete bill on this page) incl. €4.55 IPv4€8,639.80 12 mo
Contabo AMD Turin 64 Cores · St. Louis 64 cores · 128 threads · 192 GB RAM · 2 TB NVMe · Unmetered traffic Check due
64 cores 128threads 192 GB NVMe 2 TB Unmetered €725.50 €8,706.00 12 mo
Vultr Bare Metal AMD EPYC 9254 24 cores · 48 threads · 384 GB RAM · 960 GB SSD · 10.24 TB traffic +€8.91/TB Check due
24 cores 48threads DDR5 384 GB SSD 960 GB 10.24 TB +€8.91/TB +12€735.03 €8,820.39 12 mo
Scaleway Elastic Metal EM-I620E-NVMe · AMD EPYC 8534P · 2 × 3.84 TB NVMe 64 cores · 128 threads · 576 GB RAM · 7,680 GB NVMe · Unmetered traffic Check due
64 cores 128threads DDR5ECC 576 GB NVMe 7.68 TB Unmetered €736.66 incl. €56.67 setup€8,839.87 12 mo
Scaleway Dedibox Core-10-XXL · AMD EPYC 8534P · 2 × 3.84 TB NVMe 64 cores · 128 threads · 576 GB RAM · 7,680 GB NVMe · Unmetered traffic Check due
64 cores 128threads DDR5 576 GB NVMe 7.68 TB Unmetered €736.66 incl. €56.67 setup€8,839.87 12 mo
Scaleway Dedibox Store-9-XL · AMD EPYC 7313P · 6 × 3.84 TB NVMe 16 cores · 32 threads · 256 GB RAM · 23,040 GB NVMe · Unmetered traffic Check due
16 cores 32threads DDR4 256 GB NVMe 23.04 TB Unmetered €736.66 incl. €56.67 setup€8,839.87 12 mo
Akamai (Linode) G7 Dedicated 64x32 · Jakarta 32 vCPU · 64 GB RAM · 1,280 GB · 8 TB traffic +€13.36/TB Check due
Dedicated 32vCPU 64 GB 1.28 TB 8 TB +€13.36/TB €738.77 €8,865.29 12 mo
Akamai (Linode) Premium 64GB · Jakarta 32 vCPU · 64 GB RAM · 1,280 GB · 8 TB traffic +€13.36/TB Check due
Dedicated 32vCPU 64 GB 1.28 TB 8 TB +€13.36/TB €738.99 €8,867.85 12 mo
Cherry Servers Bare Metal AMD EPYC 9554P · 2 × 1 TB NVMe 64 cores · 128 threads · 384 GB RAM · 2 TB NVMe · 100 TB traffic +€0.50/TB Check due
64 cores 128threads DDR5ECC 384 GB NVMe 2 TB 100 TB +€0.50/TB €739.00 Setup fee not published€8,868.00 12 mo
Cherry Servers Bare Metal AMD EPYC 9554P · 2 × 1 TB NVMe · Šiauliai 64 cores · 128 threads · 384 GB RAM · 2 TB NVMe · 100 TB traffic +€0.50/TB Check due
64 cores 128threads DDR5ECC 384 GB NVMe 2 TB 100 TB +€0.50/TB €739.00 Setup fee not published€8,868.00 12 mo
Cherry Servers Bare Metal AMD EPYC 9554P · 2 × 1 TB NVMe · Amsterdam 64 cores · 128 threads · 384 GB RAM · 2 TB NVMe · 100 TB traffic +€0.50/TB Check due
64 cores 128threads DDR5ECC 384 GB NVMe 2 TB 100 TB +€0.50/TB €739.00 Setup fee not published€8,868.00 12 mo
Cherry Servers Bare Metal AMD EPYC 9554P · 2 × 1 TB NVMe · Singapore 64 cores · 128 threads · 384 GB RAM · 2 TB NVMe · 20 TB traffic +€1.69/TB Check due
64 cores 128threads DDR5ECC 384 GB NVMe 2 TB 20 TB +€1.69/TB €739.00 Setup fee not published€8,868.00 12 mo
AWS Lightsail Lightsail Compute Optimized 72 GB +6 locations 36 vCPU · 72 GB RAM · 1,280 GB SSD · 9 TB traffic +€80.19–€133.64/TB Check due
Dedicated 36vCPU 72 GB SSD 1.28 TB 9 TB +€80.19–€133.64/TB +10€751.96 €9,023.52 12 mo
AWS Lightsail Lightsail Compute Optimized 72 GB · Mumbai +2 locations 36 vCPU · 72 GB RAM · 1,280 GB SSD · 4.5 TB traffic +€115.82–€151.46/TB Check due
Dedicated 36vCPU 72 GB SSD 1.28 TB 4.5 TB +€115.82–€151.46/TB €751.96 €9,023.52 12 mo
phoenixNAP Bare Metal 2 × Intel Xeon Gold 5317 · 2 × 4 TB NVMe (d2.m1.large) +1 identical plan 24 cores · 48 threads · 512 GB RAM · 8 TB NVMe · 15 TB traffic +€8.91/TB Check due
24 cores 48threads 512 GB NVMe 8 TB 15 TB +€8.91/TB €752.50 incl. €4.55 IPv4€9,030.04 12 mo
phoenixNAP Bare Metal 2 × Intel Xeon Gold 5317 · 2 × 4 TB NVMe · Singapore (d2.m1.large) +1 identical plan 24 cores · 48 threads · 512 GB RAM · 8 TB NVMe · 5 TB traffic +€26.73/TB Check due
24 cores 48threads 512 GB NVMe 8 TB 5 TB +€26.73/TB €752.50 incl. €4.55 IPv4€9,030.04 12 mo
1,426–1,450 of 1,700

152 servers with IPv6 only are hidden because bills include an IPv4 address. Show them without IPv4

Some links may earn ServerDiff a commission. It never changes the order, which is the bill under the assumptions shown. How we rank

When to choose more: browser automation is the heavy part. Hermes Agent installs its own Chromium for browsing by default, and a headless browser can use more memory than the agent itself. If your agent browses the web or runs several tasks at once, or you run more than one agent on the server, raise the memory filter to 4 GB. Running a language model on the server itself is a different job: that needs a GPU server, not a VPS.

What it really costs: agent hosting is often sold as a one-click template on a VPS with a 12- to 48-month term, a low first-term price and a higher renewal price. The ranking here uses the average monthly bill over the period you choose, with the renewal price, the IPv4 address and any setup fee included, so you can see what the same server costs once the introductory term is over. The agent software is open source and installs with one script; the model provider bills you separately for what the agent uses.

Keep it safe: an agent holds API keys and can run commands on its server. Give it a server of its own, keep the system updated, and turn on backups so a bad tool run can be rolled back.