Open-source · Hosted in Europe

Your AI agent.
Your control. Our infrastructure.

Hosted Hermes agents on hardened European infrastructure. Open-source, fully configurable, with per-agent network isolation by design.

hermes — agent deploy
$ hermes agent create --name "my-assistant"
→ Provisioning isolated overlay network...
✓ Network created (net-agent-a3f2c1)
→ Deploying Hermes agent container...
✓ Agent running on node meyaru02
→ Connecting to Telegram, API server...
✓ Agent live at https://my-assistant.meyaru.host
# Your agent is ready. Fully isolated. Fully yours.
Features

Not just an agent — the entire ecosystem

We don't just host a container. We run the full stack: proxy, model routing, service discovery, dashboards, monitoring — managed by professionals.

Per-Agent Isolation

Each agent runs in its own overlay network. Agents can't see each other, can't reach the control plane. Docker-native isolation by design — no hacks, no watchers.

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.

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.

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.

Open-Source Models

Use our European inference cluster with Llama, Qwen, GLM and more. Or configure your own — any OpenAI-compatible endpoint works. Your keys, your models, your rules.

Managed by Pros

Consul service discovery, automated health checks, auto-restart on failure, CephFS redundant storage. 3-node Swarm cluster with quorum. We handle ops so you don't have to.

Isolation by design, not by config

Every agent gets its own Docker overlay network. This isn't a firewall rule that might break — it's network-level isolation baked into the infrastructure.

┌─────────────────────────────────┐
agent-A ┌──────┐ agent-B
net-A proxy net-B
10.1.0.0 10.1.0.16
└─────────────┘ └──────┘ └────────────┘
 
agent-Aagent-B: BLOCKED
agent-Aproxy: OK (LLM routing)
agent-A → DB: BLOCKED
agent-A → internet: OK (via NAT)
  • Per-agent overlay network — each agent in its own isolated network. Agents cannot communicate with each other. Period.
  • Proxy is the only crossover — the model proxy has interfaces in every agent network, but agents can only reach the proxy, not each other.
  • Control plane protected — databases, Consul, backend API: all on a separate network. Agents have zero access.
  • European data residency — all data stays on servers in Grenoble, France. GDPR-compliant.
  • No Docker socket exposure — agents run without access to the Docker daemon. No container escape vector.
Models

Open-source inference, or bring your own

Use our European GPU cluster with open-source models, or point your agent to any OpenAI-compatible endpoint. Full flexibility.

Llama 3.3 70B

Meta's flagship open model. Great for complex reasoning and multilingual tasks.

Included in Pro

Qwen 2.5 72B

Alibaba's powerful model with excellent code generation and analysis.

Included in Pro

GLM-4 9B

Zhipu's efficient model. Fast, capable, great for real-time conversations.

Included in Starter

Llama 3.2 3B

Lightweight model for simple tasks. Low latency, high throughput.

Included in Starter

Custom Endpoint

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

All plans

Auto Model Routing

Our proxy automatically selects the best model per request based on your tier.

All plans
Pricing

Simple, transparent pricing

Start free. Scale when you need. Cancel anytime. No hidden fees.

Starter

€9/month
For personal assistants and experiments
  • 1 agent
  • 2 GB RAM · 1 vCPU
  • Lightweight models (GLM-4, Llama 3.2)
  • Telegram + API access
  • 5GB storage
  • Community support
Get Started

Business

€99/month
For teams that need more
  • 10 agents
  • 16 GB RAM · 4 vCPU each
  • All models + custom endpoints
  • All channels + webhooks
  • 100GB storage
  • SLA 99.5% + dedicated support
Contact Us

Ready to deploy your agent?

Start a 14-day free trial. No credit card required.