Hermes Agent VPS 2
A starting profile for small CLI experiments with an external model API.
- Memory
- 2 GB RAM
- Shared compute
- 1 vCPU
- Storage
- 55 GB NVMe
- Bandwidth
- 2 TB
- Dedicated IP
- 1 included
- Server access
- Administrator
Hermes AgentBy Nous Research
Hermes Agent
VPS hosting.
Give good work
a memory.
Hermes Agent is an open-source AI agent from Nous Research with persistent memory, reusable skills, and a messaging gateway. Our planned Hermes Agent VPS hosting gives those workflows a private server environment. Ordering is coming soon.
View planned profiles Explore the workflows ↗
01 / PLAN YOUR WORKSPACESeven profiles.
Seven profiles.
Room for what comes next.
These planned resource profiles match our OpenClaw VPS specifications. Hermes pricing and deployment details will be published at launch.
Hermes Agent VPS 4
More room for a personal assistant, retained context, and light integrations.
- Memory
- 4 GB RAM
- Shared compute
- 2 vCPU
- Storage
- 80 GB NVMe
- Bandwidth
- 3 TB
- Dedicated IP
- 1 included
- Server access
- Administrator
Hermes Agent VPS 8
Headroom to evaluate messaging channels and browser-assisted research.
- Memory
- 8 GB RAM
- Shared compute
- 4 vCPU
- Storage
- 160 GB NVMe
- Bandwidth
- 4 TB
- Dedicated IP
- 1 included
- Server access
- Administrator
Hermes Agent VPS 16
A larger working set for overlapping tasks and supporting processes.
- Memory
- 16 GB RAM
- Shared compute
- 6 vCPU
- Storage
- 320 GB NVMe
- Bandwidth
- 5 TB
- Dedicated IP
- 1 included
- Server access
- Administrator
Hermes Agent VPS 32
Space to evaluate several separate agent profiles and larger project files.
- Memory
- 32 GB RAM
- Shared compute
- 8 vCPU
- Storage
- 640 GB NVMe
- Bandwidth
- 6 TB
- Dedicated IP
- 1 included
- Server access
- Administrator
Hermes Agent VPS 64
A high-memory profile for measured demand and substantial retained artifacts.
- Memory
- 64 GB RAM
- Shared compute
- 16 vCPU
- Storage
- 1280 GB NVMe
- Bandwidth
- 10 TB
- Dedicated IP
- 1 included
- Server access
- Administrator
Hermes Agent VPS 96
The largest planned profile for memory-heavy workspaces and data retention.
- Memory
- 96 GB RAM
- Shared compute
- 24 vCPU
- Storage
- 1600 GB NVMe
- Bandwidth
- 15 TB
- Dedicated IP
- 1 included
- Server access
- Administrator
THE RESOURCE VIEW
Compare the planned capacity.
Balance shared compute, working memory, stored files, and monthly data transfer. More RAM alone does not make a workflow reliable.
| Resource | Hermes Agent VPS 2 | Hermes Agent VPS 4 | Hermes Agent VPS 8 | Hermes Agent VPS 16 | Hermes Agent VPS 32 | Hermes Agent VPS 64 | Hermes Agent VPS 96 |
|---|---|---|---|---|---|---|---|
| Memory | 2 GB RAM | 4 GB RAM | 8 GB RAM | 16 GB RAM | 32 GB RAM | 64 GB RAM | 96 GB RAM |
| Shared compute | 1 vCPU | 2 vCPU | 4 vCPU | 6 vCPU | 8 vCPU | 16 vCPU | 24 vCPU |
| Storage | 55 GB NVMe | 80 GB NVMe | 160 GB NVMe | 320 GB NVMe | 640 GB NVMe | 1280 GB NVMe | 1600 GB NVMe |
| Bandwidth | 2 TB | 3 TB | 4 TB | 5 TB | 6 TB | 10 TB | 15 TB |
| Dedicated IP | 1 included | 1 included | 1 included | 1 included | 1 included | 1 included | 1 included |
| Server access | Administrator | Administrator | Administrator | Administrator | Administrator | Administrator | Administrator |
| Ordering | Coming soon | Coming soon | Coming soon | Coming soon | Coming soon | Coming soon | Coming soon |
02 / CONTEXT THAT OUTLASTS A CHATMemory keeps the context.
Memory keeps the context.
Skills keep the method.
Hermes brings different kinds of retained knowledge into a workflow. Knowing what each one stores helps you plan the server and the review process.
Persistent memory
Hermes can carry curated project facts and preferences between sessions. Its memory files are focused notes, with bounded capacity. They are not an unlimited record of everything you have said.
Check that important details were actually saved. Keep credentials out of project notes, and review what a fresh session recalls.
Read how Hermes memory works ↗Reusable agent skills
Skills are instructions Hermes loads for relevant work. A reviewed procedure can capture the sequence, checks, and output format for a familiar task.
Inspect generated or imported skills before relying on them. Reusing a method is different from retraining the language model or guaranteeing the same result.
Explore the Hermes skills system ↗The VPS holds the workspace.
Your selected model service handles inference when you use an external API. Keep its usage budget separate from server capacity; browsers, files, and local tools still consume VPS resources.
03 / A ROUTINE WITH A PURPOSE
Build around a useful outcome.
Start with one bounded task. Decide what success looks like, where the result should go, and which actions still need your review.
A briefing you can check.
Choose a small set of sources and an output format. Review citations and freshness, then turn the accepted procedure into a skill for the next briefing.
Plan for: browser memory, downloaded files, source review.An assistant in your channel.
The Hermes gateway connects configured services such as Telegram, Discord, and Slack. Begin with one channel and a restricted set of users before connecting more accounts.
Plan for: credentials, access rules, concurrent sessions.A recurring task with a stop point.
Hermes supports recurring and one-time scheduled work. Test a job manually, set its delivery target, and review the first scheduled result. Keep the gateway running for scheduled execution.
Plan for: service health, model budget, failure review.Read the official messaging gateway and scheduled-task guides for the integration you choose.
04 / CAPACITY FOLLOWS THE TOOLSSize the workflow,
Size the workflow,
then measure the peaks.
These are planning suggestions for an external-model setup. They are not official Hermes minimum requirements or promises about how many agents can run.
Start small.
Evaluate a simple CLI workflow at 2 GB. Consider 4 GB for a personal setup with integrations and retained state. Leave headroom for the operating system and monitor memory pressure.
Add the working tools.
Browser sessions, parsing files, and overlapping tasks can increase demand quickly. Test a representative busy period, including supporting processes and downloads.
Scale from evidence.
Use larger profiles when measurements justify more memory, compute, or retained data. Separate agent profiles need their own state; a larger server does not create application-level isolation.
These profiles include no GPU allocation. Review local-model inference separately. For network usage, see our VPS bandwidth planning guide.
05 / FIRST CHAT, THEN AUTOMATION
Get one conversation working.
Launch image and preinstallation details will be announced before ordering opens. This is the official configuration sequence after Hermes is installed on a compatible Linux system.
- 01
Choose the model connection.
Run
hermes setupand configure a supported provider. Use your own account and authentication, then review usage limits. - 02
Test in the CLI.
Run
hermesand complete a small task. Verify the answer and any saved state before enabling channels or scheduled jobs. - 03
Connect one messaging platform.
Use
hermes gateway setup. Follow the selected platform's token or pairing steps, restrict users, and test a message from that channel.
Starting from a plain server? Use the official Hermes installation guide and check its current platform requirements.
06 / KEEP CONTROL OF THE WORKSPACE
Make the useful state recoverable.
A private server gives you an operating environment to manage. It still needs a clear access policy, monitoring, and a tested recovery path.
Protect access.
Use narrow credentials and the channel's user controls. Review tools, command approvals, and the selected execution backend. A VPS boundary does not isolate every tool automatically.
Prepare secure SSH access ↗Back up the right data.
Preserve the state, configuration, and project files you need to restore. Store a protected copy away from the VPS and test recovery before depending on an automated routine.
Build an off-server backup plan ↗Keep operations explicit.
Watch disk growth, process health, API limits, and failed deliveries. Decide who maintains the OS, agent, and integrations before handing over recurring work.
Understand VPS management responsibilities ↗Hermes security documentation explains its authorization, approval, and execution boundaries.
BEFORE THE FIRST DEPLOYMENT
Hermes Agent hosting, explained.
What is Hermes Agent VPS hosting?
Hermes Agent VPS hosting is a server environment for running the Hermes assistant, its retained state, and configured tools. Hermes is the agent software; a connected model service supplies language-model inference. AwakeHost is preparing this hosting offer, and orders are not open yet.
Can I order a Hermes Agent VPS now?
Hermes Agent VPS ordering is coming soon. The seven resource profiles are a preview, and their buttons do not open checkout. Launch pricing, supported locations, and installation details will be published before orders open.
Will Hermes Agent be preinstalled?
The launch image and preinstallation details are still to be announced. The setup guidance describes the official Hermes workflow after installation; it does not promise a configured model account or messaging channel.
How much RAM does Hermes Agent need?
Capacity depends on enabled tools and concurrent work. Evaluate 2 GB for light CLI trials, 4 GB for a small personal setup, and 8 GB or more when adding browsers or overlapping tasks. These are workload-planning suggestions, not official minimums or guaranteed agent counts. Measure peak memory, CPU, and disk usage before scaling.
Does Hermes memory train the underlying AI model?
Hermes keeps curated context between sessions and can reuse skill documents for familiar work. These features do not mean that your VPS retrains the connected model. Review saved notes and skills, protect sensitive information, and back up the state you need to retain.
Can Hermes connect to Telegram, Discord, or Slack?
The official Hermes messaging gateway supports these integrations after configuration. You supply the required accounts and credentials and decide which users can reach the assistant. A supported integration is not a preconfigured channel included with hosting.
Are AI model usage and subscriptions included?
External model APIs, subscriptions, and paid tools are separate from the planned VPS resources. Choose a compatible model service, review its usage policy, and set a budget. The listed profiles include no GPU allocation; local-model inference needs a separate capacity assessment.
Is bandwidth unlimited?
The planned profiles include the finite monthly bandwidth allowances shown in the cards and comparison. Bandwidth here means data transferred per month. Customer overage terms will be confirmed at launch; no unlimited usage or unverified excess-use rate is promised.
How does this differ from OpenClaw VPS?
The planned hardware profiles match the current OpenClaw VPS resource tiers. The application workflows differ: this guide focuses on Hermes memory, skills, gateway operation, and scheduled work. OpenClaw has its own setup guide and currently published ordering options.
Coming soonYour next workspace starts
Review planned profiles Your next workspace starts
with a considered plan.
Preview the resource tiers and prepare your workflow while Hermes Agent VPS ordering is being readied.