Best VPS for Hermes Agent in 2026: Hetzner, DigitalOcean, OVH compared

Minimum specs and what happens below them

Minimum: 2 vCPU, 4GB RAM, 40GB SSD. At this spec, Hermes Agent with an external API provider (OpenRouter, Anthropic, OpenAI) runs reliably for a single user. The 4GB RAM floor is hard — at 2GB you see OOM errors on complex tasks involving browser automation or heavy file processing. The errors are not always graceful; they can corrupt the session in progress.

The 40GB storage recommendation exists because ~/.hermes/ grows over time. Session logs accumulate in ~/.hermes/sessions/ and the SQLite database (state.db) grows with every conversation. On a 20GB disk at several sessions per day, you can fill it in 6-12 months. Start with 40GB or configure log rotation early.

Docker adds overhead. If you use hermes config set terminal.backend docker — which sandbox-isolates shell commands — Docker needs an additional 1-2GB for container spin-up. On a 4GB server running the gateway and Docker simultaneously, run free -h before enabling the Docker backend. It is easy to forget it and discover the problem at 2am when a task fails.

Hetzner (community favourite, not close)

Hetzner is the near-universal community recommendation for Hermes hosting. The price-to-spec ratio beats DigitalOcean or AWS Lightsail significantly for equivalent hardware.

Prices as of April 2026. Hetzner raised CX prices approximately 8% in early 2026.

The CX23 is the minimum viable spec — adequate for a single-user instance with external API inference, Telegram gateway, and moderate task volume. The CX33 at €6.49/month is the recommended tier. The 8GB RAM headroom handles Docker backend, browser automation tasks, and occasional heavy workloads without OOM pressure. Two light Hermes instances on one CX33 is feasible with careful resource limits.

Datacenter locations: Nuremberg, Falkenstein, Helsinki (EU), Hillsboro Oregon, Ashburn Virginia (US). EU datacenters suit most European users. The US locations reduce round-trip for North American users on Telegram gateway, though the difference is modest since Telegram's own servers are distributed.

DigitalOcean, OVH, Vultr, and Hostinger

DigitalOcean is the default in most tutorials because of good documentation and a polished UI. The pricing does not compete:

At $48/month for 8GB RAM versus Hetzner's $7.30, the gap is hard to justify unless you are already running DigitalOcean infrastructure and want to co-locate.

OVH is the more interesting alternative. OVH VPS-1 (4 vCPU, 8GB RAM, 75GB SSD, unlimited traffic) sits around $6.46/month — matching Hetzner CX33 pricing with the added benefit of unmetered bandwidth, useful if your instance does heavy data retrieval. EU-hosted. Support quality is lower than Hetzner in community reports, though most self-hosters rarely need support.

Vultr pricing lands close to DigitalOcean. The High Frequency compute line (6 vCPU, 16GB, $48/month) works for multi-instance setups but the entry tier has the same pricing problem as DigitalOcean.

Hostinger KVM 2 — 2 vCPU, 8GB RAM, roughly $10-20/month depending on promo. More RAM per dollar than the Hetzner CX23 at a comparable price. Appears in a few community setup threads. Less established than Hetzner in the self-hosting community.

Raspberry Pi: does it actually work?

Raspberry Pi 5 (8GB): yes, with caveats. Community members have run Hermes Agent on Pi 5 for multi-week stretches using external API inference. The limitations are practical:

• SD card I/O under heavy SQLite writes causes periodic hangs — use NVMe via USB or an official M.2 hat
• Power outages corrupt the SQLite session database without a UPS
• ARM64 support: the Docker image supports arm64, so the Docker install works
• No useful local model inference — small quantised models (Phi-3-mini, Llama 3.2 3B) on Pi 5 run at 5-15 tokens/second on CPU alone

Pi is a legitimate option for a personal assistant that only hits external APIs. Not production-viable for anything where availability matters.

Shared VPS vs dedicated

Shared VPS (Hetzner CX series, DigitalOcean Basic) is the right call for Hermes Agent with external API inference. CPU is the bottleneck for Hermes orchestration and tool execution — not inference — and shared vCPU handles it without problem.

Dedicated or bare metal only makes sense if you run local Ollama inference on the same machine. Llama 3.1 70B on CPU needs 48+ GB RAM and 16+ physical cores to run at acceptable speed. That territory is far beyond the cost/use-case for most Hermes users. GPU VPS runs $200-500/month — at that price, Hermes OS managed hosting plus a premium API key is more economical than self-hosted GPU inference.

Multi-instance: several Hermes agents on one server

Hermes supports a profiles system (introduced in v0.6.0) for running multiple isolated instances from a single installation. Each profile gets a separate ~/.hermes-<profile>/ directory, independent memory stores and session database, isolated skill library, distinct gateway services, and separate config files.

For Docker-based multi-instance, give each container a separate volume — never run two containers against the same data directory:

Resource allocation:

Hetzner CPX32 (8GB AMD, €13.99/month) handles two instances comfortably.

Storage growth and disk management

What grows over time in ~/.hermes/:

• state.db — SQLite session database. Expect 50-200MB/year at moderate usage.
• sessions/ — JSONL session log files. Prune with hermes sessions export then delete old files.
• logs/ — gateway logs. Configure log rotation: hermes config set logging.max_size_mb 100.
• downloads/ — files created during tasks. Clean manually or set task-specific output directories.

A 40GB disk with normal single-user usage handles 2-3 years before storage becomes a concern. If your agent runs browser automation tasks that download files, use a dedicated data directory outside ~/.hermes/ from the beginning.

Total monthly cost

Running Hermes Agent on a self-hosted VPS (single user, external API inference, Telegram gateway):

Hermes OS managed hosting starts at $9.99/month — includes the server, gateway setup, automated backups, and zero maintenance. The self-hosted route is $10-25/month plus 2-4 hours of initial setup and 30-60 minutes/month of ongoing maintenance.

The math favours self-hosting on cash cost. It does not always favour it on time cost. At $50/hour developer time, a single afternoon debugging an update that broke your gateway costs more than six months of Hermes OS. That comparison changes in your favour if you run Hermes long-term with rare issues, or genuinely enjoy the infrastructure work.