Agent infrastructure
AI Agent Hosting: Compare Providers & Pricing
Compare OpenClaw VPS plans, managed agent runtimes and sandboxes. Review resources, billing conditions and official sources before choosing a provider.
Official sources Larger providers firstDefault order prioritizes large, established providers, followed by specialist platforms. This is an editorial display order, not a market-share or service-quality score.
| Company | Entry plan | Billing at a glance | Plans | Explore company |
|---|---|---|---|---|
 NorthflankSandbox Cloud coding agents in isolated environments with persistent workspaces. |
Usage-basedPay-as-you-go compute referenceCompute: $0.01667/vCPU-hour + $0.00833/GB-hour; disk, egress and models extra. | Cloud coding agents in isolated environments with persistent workspaces. Usage charges applyOwn model key: Plan dependent |
2 plansChecked 2026-10-23 | Explore |
 Fly.ioSandbox Persistent Linux Sprites for coding agents and tool execution. |
Usage-basedSprites — metered Linux executionSprites: $0.0385/CPU-hour + $0.021875/GB-hour; storage extra. | Persistent Linux Sprites for coding agents and tool execution. Usage charges applyOwn model key: Plan dependent |
1 planChecked 2026-10-23 | Explore |
 CloudzyVPS Preinstalled OpenClaw VPS with customer-controlled configuration. |
USD 2.48/moAdvertised OpenClaw starting offerRepresentative configuration; base amount is not total agent cost | Preinstalled OpenClaw VPS with customer-controlled configuration. Usage charges applyOwn model key: Plan dependent |
3 plansChecked 2026-10-23 | Explore |
 LightNodeVPS VPS guide and configurations for continuous OpenClaw execution. |
USD 7.71/moStart configurationRepresentative configuration; base amount is not total agent cost | VPS guide and configurations for continuous OpenClaw execution. Usage charges applyOwn model key: Plan dependent |
3 plansChecked 2026-10-23 | Explore |
 LettaAgent runtime Stateful agent hosting with persistent memory and separately metered tools and model usage. |
USD 20/moAPI PlanPer API subscription, plus usage | Unlimited agent creation; active-agent and tool usage billed separately. Usage charges applyOwn model key: Supported |
4 plansChecked 2026-10-23 | Explore |
 MastraAgent runtime TypeScript agent deployment, workflows and observability with CPU-based usage allowances. |
USD 0/moStarterPer platform subscription, plus resource usage | 24 CPU hours; 100K observability events; unlimited users, deployments and projects. Usage charges applyOwn model key: Plan dependent |
3 plansChecked 2026-10-23 | Explore |
How to compare these prices
Entry plans have different scopes. A personal subscription, developer API and platform workspace may be billed differently. Base fees can exclude model tokens, compute, tools, storage and network transfer. A zero base subscription does not mean unlimited free hosting. Open a company profile for all plans, full conditions and official sources.
We retain every specification, including fields the vendor has not disclosed. Entries over 30 days old are marked for review.
How to compare AI agent hosting
AI agent hosting provides the infrastructure that keeps an agent available to receive requests, run tools and retain application state. This directory compares three different choices: virtual servers you administer, managed agent runtimes, and isolated execution environments. Start with the deployment model, then compare the full bill and the resources your application needs.
OpenClaw VPS or a managed agent runtime?
An OpenClaw VPS gives you control of the operating system and installed software. Compare vCPU, RAM, NVMe or SSD storage, transfer allowance and backups. You are responsible for application updates, credentials and process supervision unless the selected service explicitly includes those tasks.
Managed runtimes handle more of the execution lifecycle. Review session limits, persistence, supported tools and regional availability. Sandboxes provide isolated environments for agent code, but conversation memory, storage retention and model access may need separate services.
Compare the complete cost, not only the entry price
A promotional VPS price, monthly platform minimum and per-second compute rate describe different things. Check the billing period, upfront commitment, renewal price and tax treatment. For metered services, estimate CPU time, allocated memory, storage and network transfer. Add model API, tool and other connected-service charges to your budget.
Which specifications matter for an agent?
Look for enough CPU and RAM for your workload, persistent storage for required state, suitable regions and networking, and documented execution limits. Then check secrets management, isolation, logs, recovery options and support scope. A server specification does not guarantee a fixed number of concurrent agents.
Explore providers and their official sources
Open the profiles for Microsoft Azure, AWS, Google Cloud or Hostinger to review plan conditions and source links. Use the table filters to compare providers by name and hosting category.
Does a zero subscription mean free agent hosting?
No. A service can have no fixed subscription while charging for compute, memory, storage or model usage. The table labels these entries as usage-based and shows published unit rates where confirmed.
How are prices and missing specifications reviewed?
Each company profile links to official product documentation and pricing sources and displays its review date. Regional offers, eligibility and capacity can change. Information that could not be confirmed remains marked as undisclosed; conflicting official figures are identified rather than silently treated as certain.