AgentOS for free on private infrastructures

[dibenedetto]

It's not clear to me if AgentOS can be used FOR FREE on a private infrastructure, that is, without any connection to the public network. As of now, I see the standard freemium business plan (free + pay for more services).

[mishramonalisha76]

Hi @dibenedetto, Only local AgentOS instances are free to connect and use with the control plane. To invite users to your org or connect to a live/private infra-deployed OS, you’ll need to upgrade to the starter plan. The starter plan includes 3 free user invites and 1 remote OS. For now, you can use the AGNOFRIENDS code to get a free upgrade during the beta test period.
Thanks

[thrugz]

Interested - where can I find the starter plan?

[xXMrNidaXx]

Self-hosting AgentOS is definitely valuable! At RevolutionAI (https://revolutionai.io) we run similar setups for enterprise clients.

Key considerations for private infra:

1. Model hosting — llama.cpp, vLLM, or TGI for local LLMs
2. Vector storage — Qdrant, Milvus, or pgvector
3. Observability — crucial for debugging agents in production
4. State management — Redis or persistent storage for agent memory

Architecture we use:

┌─────────────┐     ┌─────────────┐
│ Agent API   │────▶│ Local LLM   │
└─────────────┘     └─────────────┘
       │
       ▼
┌─────────────┐     ┌─────────────┐
│ Vector DB   │     │ Tool Server │
└─────────────┘     └─────────────┘

Benefits:

• Data stays on-prem
• No API costs after setup
• Full control over model versions
• Compliance-friendly

The main challenge is keeping up with model updates. We maintain a model registry to track tested versions.

What scale are you targeting? That affects architecture choices significantly.