Hermes infrastructure

Best VPS for Hermes Agent: Requirements and Test Candidates

The best VPS for Hermes Agent cannot be named from reliable evidence yet. Hermes documents Linux support, Python 3.11, Node.js v26, storage, state, remote backends, and local-model requirements, but it does not publish minimum RAM or CPU requirements. This guide separates those documented requirements from the VPS plans AIG is evaluating so the eventual recommendation can be based on measured resource use and recovery behavior rather than copied hosting roundups.

What Is Hermes Agent?

Hermes Agent is an open-source (MIT), self-improving AI agent by Nous Research. It creates skills autonomously from experience, persists memory across sessions via a SQLite database with FTS5 full-text search, and models users over time. The agent is primarily Python (~81%) with TypeScript (~16%) for browser automation and the WhatsApp bridge. It supports 200+ models and 15+ messaging integrations.

Hermes Agent System Requirements

Hermes Agent Installation on a VPS

Hermes provides a shell installer that provisions Python and Node.js automatically. Docker deployment is also supported via the included docker-compose.yml.

curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash

Hermes Agent Persistence and State

Hermes stores all state in ~/.hermes/:

• Configuration: ~/.hermes/config.yaml
• Secrets: ~/.hermes/.env (API keys)
• Learned skills: ~/.hermes/skills/
• Database: Local SQLite with FTS5 full-text search (memory, user modeling, cross-session continuity)
• Windows: %LOCALAPPDATA%\hermes\hermes-agent

Back up the entire ~/.hermes/ directory before upgrading or migrating to a new VPS. The SQLite database contains the agent's memory and learned skills — losing it means losing accumulated knowledge.

Hermes Agent Terminal Backends

Hermes supports 5 terminal backends that affect how commands execute on the VPS:

On a VPS, the Docker backend provides isolation between the agent's commands and the host OS. Configure with hermes config set terminal.backend.

Hermes Agent Model Providers

Hermes supports cloud models (Nous Portal, OpenAI, Anthropic, OpenRouter) and local models via Ollama custom endpoint configuration. Switch models with hermes model — no code changes needed.

For local inference, Hermes documents a recommended model (Hermes 4.3 36B at Q4_K_M quantization, ~21.8 GB, requiring a 24 GB GPU) and a smaller option (Qwen3 14B, ~9.3 GB, requiring 12-16 GB GPU). The 64,000-token context requirement limits which local models are viable.

Hermes Agent Security on a VPS

• Command approval: user must approve certain actions
• Secret management: allowlisting for credential access
• Credential storage: API keys in ~/.hermes/.env
• OAuth providers: Nous Portal, Anthropic, GitHub Copilot, xAI Grok, Qwen
• Default network exposure: not documented in primary sources — verify after installation
• Default listening ports: not documented in primary sources — verify after installation

Best VPS for Hermes Agent? Current AIG Test Candidates

Agent Infra Guide has not yet tested Hermes Agent on specific VPS plans. The providers below have documented pricing and are candidates for hands-on deployment tests. None should be treated as a tested recommendation.

AIG VPS Test Results

Equivalent Hermes Agent VPS deployment tests are pending. When complete, this section will include:

• Measured idle and active RAM consumption
• CPU utilization under representative workloads
• Cold start time from process launch to first response
• Smallest VPS plan that successfully runs the agent
• Disk space consumed by a default installation
• Recovery behavior after process kill and host reboot
• Actual ports listening after default installation