Open-source · Hosted in Europe

Your AI agent.
Your control.
Our platform.

Hosted Hermes agents on isolated European infrastructure, with the managed platform they need around them — model access, proxy, channels, voice, persistence, security and human support. You use the agent; we operate the stack.

hermes — agent deploy
$ hermes agent create --name "my-assistant"
→ Provisioning isolated environment...
✓ Secure network created
→ Deploying your Hermes agent...
✓ Agent running on server heaven-151
→ Connecting to Telegram, API server...
✓ Agent live at https://my-assistant.hermeshaven.com
# Your agent is ready. Fully isolated. Fully yours.
Features

Not just an agent — the managed platform around it

Every plan includes the isolated runtime your agent runs on, plus the platform services it depends on: managed proxy, model routing, channels, voice, monitoring, updates and Hermes-aware support.

Hosted in Europe

All infrastructure runs on servers in Grenoble, France. GDPR-compliant by design. No data leaves the EU. Your agents, your conversations, your data — stays here.

Human Support for AI Agents

Yes, really. When your AI agent misbehaves, a real person helps you fix it. DevOps and SRE engineers who know Hermes inside out — not a chatbot, not a ticket queue.

Full Control

You own your agent's config, personality, memory, and skills. Export everything anytime. No lock-in, no proprietary formats. It's open-source Hermes, unmodified.

Flexible Resources

Scale RAM and CPU per agent. From lightweight assistants to heavy research agents crunching large contexts. You pick the resources, we provision them.

Managed by Professionals

Tested Hermes and platform updates, health checks, auto-restart on failure, redundant storage. We operate the supported stack so you don't have to.

Free Open-Source Models

Qwen 3.6 35B included free on every plan, running on our European GPU cluster. Or configure your own — any OpenAI-compatible endpoint. Your keys, your models, your rules.

Free STT on GPU

Speech-to-text running on our own GPUs. Fast, private, no per-minute charges. Your agents can transcribe voice messages and audio without relying on external APIs.

SearxNG + Firecrawl

Private web search and web scraping pre-deployed and ready to use. Your agents can search the web and extract content from any page — no setup, no API keys, no extra cost.

Behind a Load Balancer

Servers are never exposed to the internet. All traffic flows through hardened proxies with SSL termination. Each agent runs in its own isolated network — no public IP, no cross-agent access. Protected by design.

Security is not a feature — it's the foundation

Your agents handle sensitive data. We built HermesHaven so that data stays yours, stays in the EU, and stays encrypted — end to end.

🌐 Internet 🔒 HTTPS Gateway Auth · TLS · Rate limiting L2 isolation — zero cross-traffic Agent A non-root Agent B non-root Agent C non-root ✕ No internet access
  • European data residency — all data stays on servers in Grenoble, France. GDPR-compliant by design. No data ever leaves the EU.
  • 2FA on every account — two-factor authentication protects your account, your agents, and your billing. No one gets in without both factors.
  • Hermes-native authentication — agent dashboards use Hermes' own auth system with token-based sessions. No third-party auth providers in the chain.
  • End-to-end encryption — TLS 1.3 in transit, AES-256 at rest. Secrets are envelope-encrypted. Your data is never plaintext outside your agent.
  • Agents isolated from the internet — your agents run behind our barrier with no public IP. They can't be reached from outside, and they can't reach each other.
  • Gateway faces the world — all traffic flows through our HTTPS gateway with TLS termination. The servers running your agents stay invisible.
  • Non-root, fully capable — agents run without root privileges, yet can still browse, search, code, and execute. Maximum capability with zero escape vectors.
Models

Included models, premium models, or your own keys

Qwen 3.6 35B is free on every plan. Premium models are available through prepaid Haven Credits. Or bring your own API keys — OpenAI, Anthropic, Ollama, any OpenAI-compatible endpoint.

Qwen 3.6 35B

Powerful open-source model. Great for general tasks, code, and conversations. Running on our own GPUs.

Free — all plans

GLM-5.2

Advanced reasoning model. Available via our proxy with pre-paid credit. No external account needed.

Via proxy — paid

GPT-5.5 (OpenAI)

OpenAI's flagship model. Buy credit from us with margin, or use your own OpenAI API key.

Proxy or BYO key

Claude (Anthropic)

Anthropic's Claude models. Available via our Haven Credits or your own Anthropic key.

Proxy or BYO key

Custom Endpoint

Any OpenAI-compatible API. Ollama, vLLM, Together, local GPU — your choice, your keys.

All plans

Haven Credits

Pre-pay €10, €50, or €100. Use any premium model. Auto-falls back to Qwen 3.6 when credits run out.

€50 → €45 · €100 → €85

Your Hermes agent, ready to run.
The platform behind it, managed.

Every plan combines isolated European hosting with the managed services Hermes needs around it: model access, secure proxying, usage controls, channels, voice, persistent data, monitoring, tested updates and human support. You use the agent; we operate the stack.

Plan differences mainly define agent capacity, usage allowances and support level. Backup frequency, premium model usage and other add-ons are shown separately.

What's actually included

A Hermes Haven plan pays for three things

The infrastructure your agent runs on, the platform services it uses, and the operations work that keeps both healthy.

Infrastructure

  • An isolated Hermes runtime

    CPU, RAM and storage assigned to the agent, with its own isolated network and no direct public server exposure.

  • Persistent agent data

    Configuration, memory, conversations, skills and working files survive restarts and routine updates. Export available according to your plan.

  • Secure public access

    TLS termination, routing and controlled ingress expose the agent through the platform rather than through a raw public IP.

Your agent has a protected place to run.

Managed platform

  • Model access and routing

    Included open models, premium models through prepaid Haven Credits, and bring-your-own OpenAI-compatible endpoints.

  • Managed AI proxy

    One controlled path for model, speech-to-text and text-to-speech providers, with credentials kept outside the agent runtime where supported.

  • Usage limits and accounting

    Plan allowances, credit or request metering, Haven Credits balance and visible usage records. You know what is included, what is metered and what happens at a limit.

  • Channels and integrations

    Supported messaging and API channels connected through guided configuration.

  • Voice services

    Speech-to-text and text-to-speech available through the managed proxy, with local or third-party processing described accurately per service.

  • Skills and configuration

    You control the agent's personality, tools, skills and configuration while Hermes Haven manages the runtime around them.

  • Backup and recovery options

    Managed backup, restore and recovery workflows with frequency and retention stated per plan or add-on.

The services around Hermes already work together.

Operations & support

  • Security and tenant isolation

    Network separation, controlled secrets, hardened ingress and account protection. Concrete controls, not generic security claims.

  • Updates and compatibility management

    Tested Hermes and platform updates, with deployment and compatibility work handled by Hermes Haven.

  • Monitoring and incident detection

    Health checks and operational monitoring for the managed runtime and platform services.

  • Recovery and human support

    Restart, repair and restore procedures plus access to people who understand Hermes and the hosting stack. Response channel and target match each plan.

Someone keeps the system healthy and helps when it fails.
Comparison

Self-managed or managed: two valid ways to use Hermes

Self-managed installation offers direct control and maximum customisation. Hermes Haven preserves Hermes configurability and reduces the day-to-day infrastructure and operations work.

Area
Self-managed Hermes on your own VPS
Managed Hermes Haven
Initial installation Guided installation; VPS preparation and administration managed by the user Automated infrastructure provisioning
Model configuration Flexible, via the Hermes wizard or configuration Providers and models configurable from the platform
Inference proxy Direct connection to providers or optional user-managed proxy Managed proxy included for services supported by the plan
Updates User decides when to apply them and manages the environment Update flow coordinated by the platform, with pre-update backup
Backups Policy, storage and restoration managed by the user Managed backups; frequency and availability per plan or add-on
Observability Tools and alerts chosen and configured by the user Operational status and metrics integrated into the platform
Security Direct control and management of the VPS, credentials and service exposure Platform controls and managed base configuration; credentials and access remain a shared responsibility
Multimedia integrations Capabilities available in Hermes; providers and credentials configured by the user Voice, transcription and images integrated per plan and available provider
Support Documentation, community or own team Platform support per plan
Time to deployment Variable depending on environment, customisation and user experience Fewer manual tasks thanks to provisioning and managed integrations
Initial installation
Self-managedHermes on your own VPS
Guided installation; VPS preparation and administration managed by the user
ManagedHermes Haven
Automated infrastructure provisioning
Model configuration
Self-managedHermes on your own VPS
Flexible, via the Hermes wizard or configuration
ManagedHermes Haven
Providers and models configurable from the platform
Inference proxy
Self-managedHermes on your own VPS
Direct connection to providers or optional user-managed proxy
ManagedHermes Haven
Managed proxy included for services supported by the plan
Updates
Self-managedHermes on your own VPS
User decides when to apply them and manages the environment
ManagedHermes Haven
Update flow coordinated by the platform, with pre-update backup
Backups
Self-managedHermes on your own VPS
Policy, storage and restoration managed by the user
ManagedHermes Haven
Managed backups; frequency and availability per plan or add-on
Observability
Self-managedHermes on your own VPS
Tools and alerts chosen and configured by the user
ManagedHermes Haven
Operational status and metrics integrated into the platform
Security
Self-managedHermes on your own VPS
Direct control and management of the VPS, credentials and service exposure
ManagedHermes Haven
Platform controls and managed base configuration; credentials and access remain a shared responsibility
Multimedia integrations
Self-managedHermes on your own VPS
Capabilities available in Hermes; providers and credentials configured by the user
ManagedHermes Haven
Voice, transcription and images integrated per plan and available provider
Support
Self-managedHermes on your own VPS
Documentation, community or own team
ManagedHermes Haven
Platform support per plan
Time to deployment
Self-managedHermes on your own VPS
Variable depending on environment, customisation and user experience
ManagedHermes Haven
Fewer manual tasks thanks to provisioning and managed integrations
The main difference is not capability: it is who handles the operations. The self-managed option prioritises direct control; Hermes Haven reduces operational time and the risk associated with manual infrastructure, update and backup tasks.
Pricing

Managed platform plans

Each plan includes the managed Hermes platform. Plan differences set your agent capacity, usage allowance and support level. 15-day free trial, no credit card. Cancel anytime.

Free Trial Starter

€0 / 15 days · no credit card
1 agent 1 GB RAM Qwen 3.6 35B 1,500 Base Credits/mes SearxNG + Firecrawl + STT Full dashboard
Sign up

Starter

€9/month
Personal assistants & experiments
  • 1 agent
  • 1 GB RAM · 1 vCPU
  • Included models · Haven Credits or BYO keys
  • 1,500 Base Credits/mes
  • 5 GB storage
  • Email support
Sign up

Pro

€39/month
Teams & multiple agents
  • 3 agents
  • 2 GB RAM · 2 vCPU each
  • Included models · Haven Credits or BYO keys
  • 20,000 Base Credits/mes
  • 20 GB storage
  • Priority support
Sign up

Enterprise

€119/month
Scale & dedicated support
  • 10 agents
  • 2 GB RAM · 2 vCPU each
  • Included models + custom endpoints
  • 50,000 Base Credits/mes
  • 40 GB storage
  • SLA 99.5% + dedicated support
Contact Us

Add-ons — scale on demand

+1 GB RAM · €5/mo +1 vCPU · €5/mo +5 GB disk · €2/mo +10 GB disk · €4/mo +30 GB disk · €10/mo +1 agent · €5/mo Hourly backups · €10/mo Haven Credits · from €10

Scale up or down anytime from your dashboard. Changes apply in seconds — no migration, no downtime.

Ready to deploy your agent?

15-day free trial. No credit card. Qwen 3.6 35B included.