Managed OpenClaw hosting: what is included, and what it really costs

A worked month: one light OpenClaw agent

List prices checked on 10 October 2026, before VAT. The token line assumes the light digest agent described above, and the hours are Qoren's estimate at an assumed $75 an hour; your workload will differ.

OpenClaw hosting comparison

What managed OpenClaw hosting includes

Managed hosting does not make tokens free; nothing does. It replaces the server and the hours. On Qoren, plans start at $39 per month, which hosts one agent all month; larger plans host more agents per environment. Each plan's monthly credits pay for that hosting and a share of managed model usage, and you can top up when you need more. You configure the agent, its schedule, and its spend policy; the operational surface is the dashboard, not SSH.

• A dedicated, isolated cloud environment, provisioned and kept online for you.
• Runtime updates handled: each environment stays on a pinned OpenClaw release that Qoren moves deliberately, so an upgrade like 2.0 is not yours to schedule.
• Secret storage, so keys and tokens never sit on your laptop or in a prompt.
• Health monitoring and alerts, with a public status page.
• Managed model key with a hard spend cap, or your own key from OpenRouter, Anthropic, OpenAI or another supported provider on every plan.
• Your existing OpenClaw skills and configuration keep working the way OpenClaw users expect.

• Compare Qoren plans
• New to OpenClaw? What it is and how it runs

The three real cost lines

Running OpenClaw around the clock has three costs, and only the first shows up on an invoice by default. Compute is the machine that keeps the agent online. Tokens are what the model provider charges for the agent's actual work, and they cost the same whether the agent lives on your laptop, a VPS, or a managed platform. Your time is setup, updates, monitoring, and fixing the agent when it stops at a bad moment, and it is the line most cost comparisons quietly leave out.

Local OpenClaw: free, until you count uptime

Running OpenClaw locally costs nothing beyond tokens and is the fastest way to experiment. It is also the easiest way to tie an agent's uptime to a machine that is designed to sleep. A closed lid, a wifi change, or a routine reboot ends the run, so local is the right choice for development and the wrong one for anything that must work overnight.

• Cost: $0 in compute, tokens only.
• Direct access to local files and tools while you build.
• Uptime ends when the laptop sleeps, reboots, or moves.

OpenClaw on a VPS: real prices

A VPS makes the agent independent of your laptop. OpenClaw's own DigitalOcean guide starts small: "Run a persistent OpenClaw Gateway on a DigitalOcean Droplet (~$6/month for the 1 GB Basic plan)." That works for a light agent once you add swap, and the guide's troubleshooting section points to a 2 GB Droplet if the agent still runs out of memory, so 2 GB is the floor we would pick for an agent that runs tools all day. As of October 2026, DigitalOcean lists a Basic droplet with 1 vCPU and 2 GB at $12 a month, and the premium Intel version of that shape with a 70 GB NVMe disk, the machine class Qoren provisions for single-agent environments, at $16. On Hetzner, the 2 GB CPX12 lists at €11.99 a month before VAT, including an IPv4 address; the cheaper cost-optimized CX23 (4 GB, €5.99) was listed as not available on 10 October 2026. What these buy is the machine only. The operating system, runtime installs, firewalling, process restarts, log rotation, and updates are yours, and so is noticing when the agent dies.

• DigitalOcean Basic, 1 vCPU / 1 GB: $6/mo list, the size OpenClaw's guide uses, with swap added.
• DigitalOcean Basic, 1 vCPU / 2 GB: $12/mo list.
• DigitalOcean premium Intel, 1 vCPU / 2 GB / 70 GB NVMe: $16/mo list, the class Qoren uses for single-agent environments.
• Hetzner CPX12, 1 vCPU / 2 GB: €11.99/mo list before VAT, including an IPv4 address.
• Bigger agents or several agents per host push you to 2 vCPU / 4 GB: $24/mo for the DigitalOcean Basic shape, €19.99 for Hetzner's CPX22.

Sources: DigitalOcean Droplet pricing · Hetzner Cloud: Regular Performance pricing · Hetzner Cloud: Cost-Optimized pricing · OpenClaw docs: DigitalOcean

Tokens: two worked examples

The model bill is driven by how much the agent reads and writes, not where it runs, so the same numbers apply to every hosting path. A light monitoring agent that runs three times a day, reading about 40,000 tokens and writing about 4,000 per run on a model priced at $1 per million input tokens and $5 per million output tokens, such as Claude Haiku 4.5 on OpenRouter as of October 2026, costs about six cents a run, or 18 cents a day: roughly $5-6 a month. A heavier agent doing one substantial task a day, around 200,000 tokens in and 20,000 out on a model at $3 and $15 per million, such as Claude Sonnet 4.6, is about 90 cents a day, or roughly $27 a month. Busy agents can spend much more, which is why a hard cap matters more than the exact rate.

• Light digest agent, 3 runs/day: ≈ $5-6/mo in tokens.
• Heavier daily task agent: ≈ $27/mo in tokens.
• Identical on self-hosted and managed; cap it either way.
• On Qoren, spend draws from your monthly credit grant, then any topped-up credits, and stops at your cap.

Sources: OpenRouter: Claude Haiku 4.5 pricing · OpenRouter: Claude Sonnet 4.6 pricing

Your time: the biggest line for most people

By Qoren's estimate, getting a hardened host from zero to a live agent takes 2-4 hours the first time: provisioning, locking down SSH, installing the runtime, wiring secrets, and testing restarts. That figure adds up those steps for someone doing them for the first time; it is not a measurement, and OpenClaw's own DigitalOcean guide budgets about 20 minutes for the basic install, before any hardening or alerting. Our estimate for upkeep is 1-3 hours a month across updates, log checks, and the occasional stalled process, plus a bad month now and then when a dependency breaks. At an assumed $75 for your working hour, that upkeep alone is $75-225 a month in time, several times the server bill. Self-hosting is not expensive because of the VPS; it is expensive because you become the on-call engineer.

Sources: OpenClaw docs: DigitalOcean

Uptime: the number nobody publishes for a laptop

Provider SLAs cover the virtual machine, not your agent. A VPS can be up while the agent inside it has been dead for days, because process crashes, full disks, and failed updates are outside the provider's contract, and nothing alerts you unless you built the alerting. On Qoren, environments are health-monitored, faults raise alerts you can see in the dashboard, and service status for the last 30 days is public.

Sources: Qoren service status

When self-hosting is genuinely cheaper

If you already operate servers, patching and monitoring are marginal work you have absorbed, and one more process on an existing host is close to free. Self-hosting also wins when you need host-level control or a locked-down private network as part of the design. The honest rule: if the hours are already spent, self-host; if the hours would be new, they are the cost.