Agents
Managed instances of open-source agent frameworks, billed per hour. Plans are synced from provider catalogs.
Agents price list
| Agent | Plan | vCPU | RAM | $/hour | Provider | Last sync | |
|---|---|---|---|---|---|---|---|
HR
Harness Runtime: Large
|
mars-4vcpu-8gb |
4 | 8 GB | — | $0.252 | DigitalOcean | |
|
HR
Harness Runtime: Medium
|
mars-2vcpu-4gb |
2 | 4 GB | — | $0.126 | DigitalOcean | |
|
HR
Harness Runtime: Small
|
mars-2vcpu-2gb |
2 | 2 GB | — | $0.107 | DigitalOcean | |
|
HR
Harness Runtime: XLarge
|
mars-16vcpu-32gb |
16 | 32 GB | — | $1.008 | DigitalOcean | |
|
HR
Harness Runtime: XSmall
|
mars-1vcpu-1gb |
1 | 1 GB | — | $0.0535 | DigitalOcean | |
HA
Hermes-Agent
|
tiny |
2 | 1 GB | 2048 GB | $0.018 | DeepInfra | |
|
HA
Hermes-Agent
|
budget |
2 | 2 GB | 3072 GB | $0.03 | DeepInfra | |
O
OpenClaw
|
budget |
2 | 2 GB | 3072 GB | $0.03 | DeepInfra | |
|
O
OpenClaw
|
standard |
2 | 4 GB | 4096 GB | $0.05 | DeepInfra |
Plans reflect the last sync (14h ago). Hourly billing for always-on agent instances.
Frequently asked questions
What is agent hosting?
Always-on cloud instances sized for AI agents and their tools. Your agent keeps its memory, browser sessions and background tasks alive between messages — something serverless inference cannot do.
How are agent plans billed?
Per hour of uptime in USD, like dedicated hardware. The vCPU, RAM and monthly egress allowance in each row come from the provider's own plan catalog.
How do I choose a plan?
Match the plan to your agent's tool load: heavy browser or code-execution workloads need more vCPU and RAM, while a mostly-API agent can run on the smallest tier. Compare the hourly rates below across providers.