Hermes WebUI

Hermes WebUI is a browser interface for Hermes Agent, the self-hosted autonomous agent from Nous Research that keeps memory across sessions, runs scheduled jobs, and is reachable from messaging apps. The web app aims for parity with the Hermes command-line experience and runs the agent in-process, reading your existing HERMES_HOME configuration and model providers rather than requiring separate setup.

The server is Python with a vanilla JavaScript frontend, with no build step or bundler. The layout has three panels: a left sidebar for sessions and navigation, a central chat view, and a right-hand workspace file browser. Model, profile, and workspace controls live in the composer footer, and a context ring shows token usage. Settings and session tools are collected in a Hermes Control Center.

You typically run it on a machine you control and reach it through an SSH tunnel, Tailscale, or a native client on another device, so one server holds your conversations, memory, scheduled jobs, and workspace. A ctl.sh helper manages it as a daemon, and Nix flake, NixOS module, and Docker deployments are available.

Features

• Streaming chat: SSE token streaming with auto-reconnect, message queuing, inline edit and regenerate, retry, and cancel
• Tool and subagent cards: inline cards for tool calls and child-agent activity, plus collapsible reasoning blocks and approval prompts for dangerous shell commands
• Providers: OpenAI, Anthropic, Google, DeepSeek, Nous Portal, OpenRouter, MiniMax, Xiaomi MiMo, Z.AI, and custom OpenAI-compatible endpoints such as Ollama or LM Studio
• Sessions: create, rename, pin, archive, tag, and group into projects; export to Markdown or JSON; public read-only share links
• CLI session bridge: sessions from the Hermes CLI SQLite store appear in the sidebar and can be imported with full history
• Workspace browser: directory tree, inline preview of text, code, Markdown, and images, file editing and creation, and git branch and dirty-file badges
• Voice input: Web Speech API dictation with live interim transcription
• Profiles: switch between agent profiles without a restart, with gateway status and per-profile model info
• Rendering: Mermaid diagrams, Prism.js syntax highlighting, and code copy buttons
• Cost tracking: input and output token counts and estimated cost per conversation