Hermes Agent

Hermes Agent is an MIT-licensed AI agent from Nous Research. It combines CLI, terminal and Desktop interfaces with saved skills, conversation history, memory and scheduled work. It supports multiple model providers and compatible endpoints. Software licensing does not include free inference or hosting.

Continuing work

The project documents a learning loop that creates and updates reusable skills. Version 0.23.0 adds browser access to the Desktop workspace through the user's gateway; Webapp access is off by default. One gateway owns each session and interfaces connect to the same conversation. Rewind controls let you edit/resend an earlier message or branch while preserving the original.

Installation and platforms

• Windows: native installation is supported; Desktop uses an App Installer/MSIX route.
• macOS: Desktop requires macOS 13+ and Apple Silicon. Intel Macs are not supported according to the installation guide.
• Linux: use the official installer/source route or documented container options. Do not expect the identical Desktop package offered for Windows or macOS.
• Browser: optional Webapp served by a configured gateway, not a hosted account that works without setup.

Cost, data and permissions

The code is free under MIT. Model API usage, hosting and optional services can cost money. Running the agent on your computer does not keep requests local if you select a remote provider. Review model routes, tool connections and permissions before supplying sensitive files. For command/file work, begin in an isolated or disposable project and review saved skills rather than assuming every learned behavior is correct.

Screenshot: real official model-settings dashboard asset marked v0.11.0, used as a historical interface illustration. It is not a capture of the new 0.23.0 Webapp.

Read our Hermes Agent discovery. Sources: repository, release 0.23.0, installation guide, MIT license. Documentation-based overview, not a reliability or security test.