Hermes Agent

Hermes Agent is an open-source AI agent from Nous Research for task execution from a terminal, native desktop app, messaging gateway, or supported editor workflow. It combines an interactive agent with persistent user and project context, reusable skills, scheduled jobs, subagent delegation, web and media tools, and Model Context Protocol connections. Users select a supported cloud provider, Nous Portal, or an OpenAI-compatible endpoint and can route work to local, container, remote, or hosted execution backends. The public repository identifies the project as MIT-licensed.

Quick Setup uses Nous Portal OAuth, while Full Setup supports bring-your-own provider credentials. Provider or hosted-service charges can vary; the agent software itself does not imply that every model or connected tool is free. Deployment prerequisites depend on the selected operating system and runtime.

Top features

• Terminal and desktop interfaces: use an interactive TUI or a native desktop surface while retaining tool output, conversation history, and project context.
• Messaging gateway: connect one agent runtime to supported chat and communication channels and route replies back to the originating platform.
• Persistent memory and session search: retain user preferences and project knowledge, then retrieve earlier work across conversations.
• Skills and extensions: package repeatable procedures as skills and connect external capabilities through built-in tools, plugins, and MCP servers.
• Scheduled work: cron jobs can launch recurring reports, checks, backups, or briefings and deliver the result through a configured channel.
• Delegation and pipelines: isolated subagents can handle parallel work, while Python RPC scripts can orchestrate tool calls with less main-session context use.
• Execution choices: documented backends include local, Docker, SSH, Singularity, Modal, Daytona, and Vercel Sandbox paths.

Use cases

• Run a personal assistant on a workstation or server and reach it from the terminal or a configured messaging service.
• Delegate repository research, documentation, testing, or operational checks to bounded subagents.
• Schedule a recurring digest or audit and send the completed artifact to a chosen channel.
• Connect internal tools and data through MCP while keeping project instructions in reusable skills.
• Use remote or isolated execution when a task should not run directly on the local host.