Hermes Agent is a self-hosted AI agent published by Nous Research. It keeps persistent memory, runs scheduled jobs, and stays alive on your own machine or VPS — but its standard interface is a terminal CLI. Hermes Web UI is an unofficial, community-built web interface that lets you drive that same agent from a browser. It is developed by Nathan Esquenazi (GitHub: nesquena), and the official site states plainly that it is not affiliated with or endorsed by Nous Research. A three-pane layout of session list, chat, and workspace file browser gives you roughly the same reach as the CLI, from a desktop browser or a phone. It is open source under the MIT license.
Key Features
- Three-pane layout with CLI-level parity: Session list on the left, chat in the middle, workspace file browser on the right. You can preview, edit, create, and delete files, and the browser detects Git-tracked directories and lets you drag panel widths
- Streaming responses and tool-call cards: Replies stream in token by token. Tool calls the agent makes are collapsed into cards you can expand to inspect, and the chat renders inline Mermaid diagrams and reasoning blocks
- Session organization and sharing: Group sessions by project or tag, search them, archive them, download them as Markdown or JSON, and issue public share links
- Scheduled task (cron) management: List, create, run, and manage the agent’s recurring jobs from the same screen
- Mobile support and voice input: A responsive hamburger sidebar with touch-sized targets, PWA installation on supported browsers, and dictation via the Web Speech API microphone button
- Profile switching and authentication options: Switch agent profiles without a restart, and protect access with password auth, passkeys (WebAuthn), OIDC login, SSH tunneling, or a reverse proxy
Pricing
| Plan | Price | Highlights |
|---|---|---|
| Self-hosted | $0 | Open source under the MIT license, with no feature limits |
| Additional running costs | Usage-based | API charges from your LLM provider, plus server or VPS costs |
Pricing is as of August 2026. The official site also points to third-party managed hosting (from $3.99/month), but the software itself is free to self-host. Check the official site for the latest details.
Pros & Cons
✅ Pros
- Send instructions to the agent from a phone or tablet without opening a terminal
- Inspect and edit workspace files on the same screen, so it is easier to see what the agent actually produced
- Free under the MIT license, with conversations and files staying on your own server
- A full set of authentication options for exposing the UI beyond localhost
- Installs via Docker Compose or a bootstrap script, with no Node.js build step (it runs on the Python standard library and plain JavaScript)
⚠️ Cons
- It is an unofficial project and is not covered by Nous Research support
- It remains tightly coupled to Hermes Agent internals, and a stable API boundary is still open work
- There is no native Windows installer; you need WSL2
- In multi-container Docker setups, tools run in the Web UI container rather than the agent container, so behavior varies by configuration
- It assumes Python 3.11+ and someone willing to administer a server, so it is not for complete non-engineers
Comparison with Similar Services
| Criteria | Hermes Web UI | Hermes Agent CLI | Open WebUI | LibreChat |
|---|---|---|---|---|
| Positioning | Unofficial web UI for Hermes Agent | Hermes Agent’s standard terminal interface | Self-hosted general-purpose chat UI | Self-hosted general-purpose chat UI |
| Connects to | Hermes Agent | Hermes Agent | Various LLMs and local models | Various LLM APIs |
| Phone access | Browser / PWA | Not available | Browser | Browser |
| Scheduled task management | In the UI | Via commands | ─ | ─ |
| Workspace file operations | Built-in file browser | Via the shell | ─ | ─ |
| Cost | Free (self-hosted) | Free | Free (self-hosted) | Free (self-hosted) |
Who Is It For
- People running Hermes Agent around the clock on a VPS or home server who want to check progress while away
- People who would rather survey sessions and workspaces in a GUI than in a terminal
- People who want to hand agent output to a teammate or friend through a share link
- People who want to dictate instructions to the agent from a phone while on the move
- People who would rather keep conversation history and generated files under their own control than in an external SaaS
Summary
Hermes Web UI extends a terminal-first agent to the browser and the phone. The three-pane layout puts sessions, chat, and files in one view, and the authentication and remote-access options are broad enough for everyday use of a self-hosted agent. The trade-offs are worth stating up front: it is unofficial, tightly coupled to Hermes Agent internals, and it assumes you can run a server. Start it locally first, and if it earns a place in your routine, look at reaching it safely from outside through something like Tailscale or Cloudflare Tunnel.