On file for Hermes (Nous Research)
- Paid · Open source core; hosted portal plans billed separately
- Operators who want an agent that runs on their own hardware with their own API keys, rather than a rented seat
- AI Agents & Workflow Builders
- Written review
What is Hermes Agent (Nous Research)?
Hermes Agent is an open-source, self-hosted autonomous agent runtime built by Nous Research. It is not a SaaS product. You clone the repository, deploy it via Docker on your own hardware or any cloud VM, and the runtime executes tasks with full access to the underlying machine — shell commands, headless browser sessions, scheduled jobs, and persistent memory.
The problem it solves is straightforward: closed agent products like Devin execute inside vendor clouds, bill per compute unit, and never touch your actual infrastructure. Hermes Agent inverts that model. The runtime is free under a permissive open-source license, your task history and memory vectors live in your own storage, and model calls go through your own API keys (BYOK).
Architecturally, it separates into four layers:
- Execution sandbox — full system terminal access plus browser automation against a headless browser instance
- Orchestration — a kanban-style task state machine (backlog → in-progress → blocked → review → done) and a native cron scheduler for recurring jobs
- Memory — Mem0-backed persistent memory that survives restarts and sessions, storing task outcomes, environment context, and learned patterns
- Model router — provider-agnostic BYOK routing across Claude, GPT, DeepSeek, GLM, and local endpoints via Ollama or vLLM
Deployment targets are local machines (macOS, Linux, WSL2) or any commodity cloud VM. There is zero data lock-in: pull the container, point it at new model keys, and migrate.
Who is Hermes Agent (Nous Research) for?
Benefits most:
- Platform and DevOps engineers who want an agent with shell reach into their own servers, CI systems, and monitoring stacks — not a sandboxed vendor environment
- Teams with data residency or compliance constraints — with local models via Ollama/vLLM, nothing leaves your network at all
- Operators running recurring jobs — log triage, report pulls, uptime checks, ETL glue — since cron scheduling is native, not a premium add-on
- **Cost
Henryk Wexel runs Quota Engine. He spent a decade selling B2B software and building outbound systems — as a founder, as a seller at Silverfort and Cradlepoint, and as the person on the hook for pipeline numbers. He writes the reviews here because he has bought, deployed, argued about and cancelled most of the tools in this index. Based in Palma de Mallorca.