Frequently Asked Questions About Hermes Agent
Find answers to common questions about installing, configuring, and running Hermes Agent across platforms, providers, integrations, memory, and automation workflows.
Hermes Agent is a self-improving AI agent by Nous Research. It combines tool use, persistent memory, skills, messaging integrations, and automation rather than stopping at chat alone.
The recommended path is the official installer script. After installation, reload your shell and start with `hermes`, `hermes model`, or `hermes setup` depending on how much configuration you want to complete up front.
Official guidance centers on Linux, macOS, WSL2, and Termux on Android. Hermes can also reach many messaging platforms through its gateway layer.
Hermes is provider-agnostic. Common paths include Nous Portal, OpenRouter, OpenAI, and other compatible endpoints exposed through runtime configuration.
Hermes is designed to run where you choose. Privacy and control depend on where you host it and which model providers or MCP servers you connect.
Skills package repeatable workflows and MCP connects Hermes to external tool servers. Together they let you move from a general agent to a domain-shaped one.
Hermes is built to live as a long-running agent. It can remember, schedule work, use tools, drive browsers, run terminal workflows, connect to messaging platforms, and improve through skills.
The software itself is open source. Your real costs come from model APIs, hosted infrastructure, storage, and whichever external services you choose to wire in.
Yes. Hermes includes migration flows for users coming from OpenClaw, including settings, memories, skills, and some allowlisted secrets.
Hermes is best suited to developers, operators, researchers, and technical users who want an agent that can run, remember, and improve over time.