A server for an AI agent

1,852 servers from 53 providers, from €4.49 a month. Prices checked 12 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
+€20.83/mo(31% 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 €20.83 a month. The list is always sorted by the full bill.)
Bill over
Bill horizon

1,852 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
AWS Lightsail Lightsail 128 GB +6 locations 32 vCPU · 128 GB RAM · 1,280 GB SSD · 9 TB traffic +€80.19–€133.64/TB Check due
32vCPU 128 GB SSD 1.28 TB 9 TB +€80.19–€133.64/TB +10€787.60(lowest complete bill on this page) €9,451.18 12 mo
AWS Lightsail Lightsail 128 GB · Mumbai +2 locations 32 vCPU · 128 GB RAM · 1,280 GB SSD · 4.5 TB traffic +€115.82–€151.46/TB Check due
32vCPU 128 GB SSD 1.28 TB 4.5 TB +€115.82–€151.46/TB €787.60(lowest complete bill on this page) €9,451.18 12 mo
Vultr VX1 General Purpose 32 vCPU · 128 GB RAM · 1,920 GB NVMe · 9.22 TB traffic +€8.91/TB Check due
Dedicated 32vCPU 128 GB NVMe 1.92 TB 9.22 TB +€8.91/TB €795.43 €9,545.15 12 mo
phoenixNAP Bare Metal 2 × Intel Xeon Gold 6336Y · 2 × 2 TB NVMe (d2.c4.large) 48 cores · 96 threads · 256 GB RAM · 4 TB NVMe · 15 TB traffic +€8.91/TB Check due
48 cores 96threads 256 GB NVMe 4 TB 15 TB +€8.91/TB €799.98 €9,599.79 12 mo
phoenixNAP Bare Metal 2 × Intel Xeon Gold 6336Y · 2 × 2 TB NVMe · Singapore (d2.c4.large) 48 cores · 96 threads · 256 GB RAM · 4 TB NVMe · 5 TB traffic +€26.73/TB Check due
48 cores 96threads 256 GB NVMe 4 TB 5 TB +€26.73/TB €799.98 €9,599.79 12 mo
BandwagonHost SPECIAL 1280G KVM PROMO V5 - CN2 GIA ECOMMERCE HIBW 20T 12 vCPU · 64 GB RAM · 1,280 GB · 20 TB traffic Check due
12vCPU 64 GB 1.28 TB 20 TB +1€800.96 Setup fee not published€9,611.55 12 mo
Akamai (Linode) Linode 128GB · Jakarta 24 vCPU · 128 GB RAM · 2,560 GB SSD · 20 TB traffic +€13.36/TB Check due
24vCPU 128 GB SSD 2.56 TB 20 TB +€13.36/TB €821.10 €9,853.17 12 mo
Akamai (Linode) G8 Dedicated 96x24 24 vCPU · 96 GB RAM · 983 GB · 0 TB traffic +€4.45/TB Check due
Dedicated 24vCPU 96 GB 983 GB 0 TB +€4.45/TB +2€823.95 incl. €4.46 Traffic€9,887.39 12 mo
phoenixNAP Bare Metal 2 × Intel Xeon Gold 6536 · 2 × 2 TB NVMe (d3.c1.large) 32 cores · 64 threads · 256 GB RAM · 4 TB NVMe · 15 TB traffic +€8.91/TB Check due
32 cores 64threads 256 GB NVMe 4 TB 15 TB +€8.91/TB €826.00 €9,911.97 12 mo
phoenixNAP Bare Metal 2 × Intel Xeon Gold 6536 · 2 × 2 TB NVMe · Singapore (d3.c1.large) 32 cores · 64 threads · 256 GB RAM · 4 TB NVMe · 5 TB traffic +€26.73/TB Check due
32 cores 64threads 256 GB NVMe 4 TB 5 TB +€26.73/TB €826.00 €9,911.97 12 mo
HostDime 2 × Intel Xeon E5-2683 V4 · Orlando 32 cores · 64 threads · 96 GB RAM · 480 GB SSD · 25 TB traffic Check due
32 cores 64threads 96 GB SSD 480 GB 25 TB €827.69 Setup fee not published€9,932.29 12 mo
Vultr Storage Optimized · São Paulo 8 vCPU · 64 GB RAM · 1,920 GB NVMe · 8.19 TB traffic +€8.91/TB Check due
Dedicated 8vCPU 64 GB NVMe 1.92 TB 8.19 TB +€8.91/TB €828.58 €9,942.98 12 mo
phoenixNAP Bare Metal Intel Xeon 6527P · 2 × 4 TB NVMe (s5.x6.c3.large) 24 cores · 48 threads · 256 GB RAM · 8 TB NVMe · 15 TB traffic +€8.91/TB Check due
24 cores 48threads 256 GB NVMe 8 TB 15 TB +€8.91/TB €839.01 €10,068.07 12 mo
phoenixNAP Bare Metal Intel Xeon 6527P · 2 × 4 TB NVMe · Singapore (s5.x6.c3.large) 24 cores · 48 threads · 256 GB RAM · 8 TB NVMe · 5 TB traffic +€26.73/TB Check due
24 cores 48threads 256 GB NVMe 8 TB 5 TB +€26.73/TB €839.01 €10,068.07 12 mo
Vultr General Purpose 32 vCPU · 128 GB RAM · 640 GB NVMe · 9.22 TB traffic +€8.91/TB Check due
Dedicated 32vCPU 128 GB NVMe 640 GB 9.22 TB +€8.91/TB +10€855.31 €10,263.72 12 mo
Vultr Memory Optimized 24 vCPU · 192 GB RAM · 1,200 GB NVMe · 11.26 TB traffic +€8.91/TB Check due
Dedicated 24vCPU 192 GB NVMe 1.2 TB 11.26 TB +€8.91/TB +11€855.31 €10,263.72 12 mo
Akamai (Linode) High Memory 300GB 16 vCPU · 300 GB RAM · 340 GB · 9 TB traffic +€4.45/TB Check due
Dedicated 16vCPU 300 GB 340 GB 9 TB +€4.45/TB +9€855.31 €10,263.72 12 mo
Akamai (Linode) G7 Dedicated 64x32 · Sao Paulo 32 vCPU · 64 GB RAM · 1,280 GB · 8 TB traffic +€6.24/TB Check due
Dedicated 32vCPU 64 GB 1.28 TB 8 TB +€6.24/TB €861.90 €10,342.84 12 mo
1,576–1,600 of 1,852

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.