Hermes Agent is a self-improving AI agent built by Nous Research, officially positioned as "the agent that grows with you." Its core feature is a built-in learning loop: it can create skills from experience, improve skills during use, proactively prompt itself to persist knowledge, search its own past conversations, and gradually build a deep understanding model of the user across sessions.
In terms of deployment, Hermes Agent can run on a VPS costing as little as $5, a GPU cluster, or serverless infrastructure that incurs almost no cost when idle. It is not tied to the user's laptop, and users can interact with the agent running on a cloud VM through messaging platforms such as Telegram.
In terms of model selection, Hermes Agent does not lock users into any single provider. Users can use Nous Portal, OpenRouter, OpenAI, their own endpoint, and many other model services, switching via the hermes model command without modifying code.
Platform support includes Linux, macOS, WSL2, and native Windows (PowerShell). Native Windows can run the CLI, gateway, TUI, and tools without WSL; if WSL2 is preferred, the one-line installation command for Linux/macOS also applies. For Android/Termux, the official signed APT repository is provided for aarch64 devices, including a stable channel (tagged releases) and a pre-release canary channel. The package includes Python, Node.js, and the TUI, and requires the Termux guide rather than the desktop/server installation script.
After installation, users can launch the terminal UI via hermes, or run the gateway and then converse with it from Telegram, Discord, Slack, WhatsApp, Signal, or Email. The two entry points share many slash commands. Full documentation is located at hermes-agent.nousresearch.com/docs.
Hermes Agent also provides Nous Portal integration, which avoids separately collecting multiple API keys for models, web search, image generation, TTS, and cloud browser, etc. Nous Portal covers 300+ models with a single subscription (selectable via /model <name>), and provides a tool gateway: web search, image generation (FAL), text-to-speech (OpenAI), and cloud browser (Browser Use) are all routed through the subscription, with no additional accounts required. Users can still bring their own keys per tool at any time; the gateway is enabled per backend rather than all-or-nothing.
For users migrating from OpenClaw, Hermes can automatically import settings, memory, skills, and API keys. During first-time setup, the hermes setup wizard automatically detects ~/.openclaw and offers a migration option before configuration begins; migration can also be done at any time after installation. Imported content includes the SOUL.md persona file, MEMORY.md and USER.md memory entries, user-created skills (imported to ~/.hermes/skills/openclaw-imports/), command allowlists, messaging platform configurations, API keys (Telegram, OpenRouter, OpenAI, Anthropic, ElevenLabs), TTS assets, and the workspace instruction AGENTS.md.
The project uses the MIT license, is built by Nous Research, and welcomes community contributions. There are also community extensions, such as computer-use-linux (a Linux desktop control MCP server) and HermesClaw (a WeChat bridge).
Regarding details such as fees, pricing, and specific performance results, complete information is not currently available in the official materials.
