Hermes Desktop Companion

Hermes Desktop Companion

Skill by ara.so — Hermes Skills collection.

Hermes Desktop is a native Electron-based desktop application for installing, configuring, and interacting with Hermes Agent — a self-improving AI assistant with tool use, multi-platform messaging, and closed learning loop. It provides a GUI alternative to CLI management with streaming chat, session management, profile switching, skill installation, memory editing, and gateway configuration.

What It Does

• Guided installation: First-run wizard installs Hermes Agent to ~/.hermes with dependency resolution
• Local or remote mode: Run Hermes locally on 127.0.0.1:8642 or connect to remote API server
• Multi-provider support: OpenRouter, Anthropic, OpenAI, Google Gemini, xAI Grok, Nous Portal, Qwen, MiniMax, Hugging Face, Groq, and local OpenAI-compatible endpoints
• Streaming chat UI: SSE-based real-time chat with tool progress, markdown rendering, syntax highlighting, token usage tracking
• 22 slash commands: /new, /clear, /fast, /web, /image, /browse, /code, /shell, /usage, /help, /tools, /skills, /model, /memory, /persona, /version, /compact, /compress, /undo, /retry, /debug, /status
• Profile management: Multiple isolated Hermes environments with separate configs
• 14 toolsets: Web search, browser automation, terminal, file ops, code execution, vision, image gen, TTS, skills, memory, session search, clarify, delegation, MoA, task planning
• Memory system: View/edit entries, user profile memory, capacity tracking, multiple providers (Honcho, Hindsight, Mem0, RetainDB, Supermemory, ByteRover)
• 16 messaging gateways: Telegram, Discord, Slack, WhatsApp, Signal, Matrix, Mattermost, Email, SMS, iMessage, DingTalk, Feishu, WeCom, WeChat, Webhooks, Home Assistant
• Session search: Full-text search (SQLite FTS5) across conversation history
• Scheduled tasks: Cron job builder with 15 delivery targets
• Hermes Office (Claw3d): Visual 3D interface with adapter management

Installation

Download Pre-built Binaries

Download from GitHub Releases:

macOS Installation

# After installing the .dmg, remove quarantine attribute
xattr -cr "/Applications/Hermes Agent.app"

Or right-click → Open → confirm in dialog.

Windows Installation

Download .exe from releases. Windows SmartScreen will warn (unsigned) — click "More info" → "Run anyway".

Future winget support:

winget install NousResearch.HermesDesktop

Linux Installation

AppImage (universal):

chmod +x hermes-desktop-*.AppImage
./hermes-desktop-*.AppImage

Debian/Ubuntu:

sudo dpkg -i hermes-desktop-*.deb
sudo apt-get install -f  # Fix dependencies if needed

Fedora/RHEL:

sudo dnf install ./hermes-desktop-*.rpm --nogpgcheck

WSL Passwordless Sudo (if installer stalls)

echo "$USER ALL=(ALL) NOPASSWD: ALL" | sudo tee /etc/sudoers.d/hermes-install
# Re-run installer, then:
sudo rm /etc/sudoers.d/hermes-install

First-Run Setup

1. Choose deployment mode:

   • Local: Installs Hermes to ~/.hermes, runs on 127.0.0.1:8642
   • Remote: Connect to existing Hermes API server (requires URL + API key)

2. Local mode: Installer checks for Git, uv, Python 3.11+, then runs official Hermes install script

3. Configure provider: Select from OpenRouter, Anthropic, OpenAI, Google, xAI, Nous Portal, Qwen, MiniMax, Hugging Face, Groq, or custom local endpoint

4. Enter credentials: API keys are saved to Hermes config files (not stored by desktop app)

5. Launch workspace: Main UI opens with Chat, Sessions, Agents, Skills, Memory, Tools, etc.

Configuration

Provider Configuration

Settings → Provider Config:

// Example: OpenRouter setup
{
  provider: "openrouter",
  apiKey: process.env.OPENROUTER_API_KEY,
  model: "anthropic/claude-3.5-sonnet",
  baseUrl: "https://openrouter.ai/api/v1"
}

// Example: Local Ollama
{
  provider: "openai",
  apiKey: "not-needed",
  model: "llama3.1:8b",
  baseUrl: "http://localhost:11434/v1"
}

Remote Mode Configuration

First run:

• Choose "Remote" mode
• Enter API URL: https://your-hermes-server.com
• Enter API key (from remote Hermes /api/keys)
• App validates connection before proceeding

Profile (Agent) Management

Agents screen:

• Create new profile: Creates isolated ~/.hermes/profiles/<name> directory
• Switch profile: Restarts Hermes with new profile context
• Delete profile: Removes profile directory and data

# Profiles stored at:
~/.hermes/profiles/
  ├── default/
  │   ├── config.yaml
  │   ├── SOUL.md
  │   ├── memory.db
  │   └── sessions.db
  └── work/
      └── ...

Memory Provider Configuration

Memory screen → Configure Providers:

// Example: Honcho
{
  provider: "honcho",
  apiKey: process.env.HONCHO_API_KEY,
  appId: "hermes-desktop",
  userId: "user-123"
}

// Example: Mem0
{
  provider: "mem0",
  apiKey: process.env.MEM0_API_KEY,
  userId: "user-123"
}

Gateway Configuration

Gateway screen → Select platform → Configure:

Telegram:

enabled: true
bot_token: ${TELEGRAM_BOT_TOKEN}
allowed_users: [123456789]

Discord:

enabled: true
token: ${DISCORD_BOT_TOKEN}
channel_ids: [1234567890123456789]

Email (IMAP/SMTP):

enabled: true
imap_server: imap.gmail.com
imap_port: 993
smtp_server: smtp.gmail.com
smtp_port: 587
username: ${EMAIL_USERNAME}
password: ${EMAIL_APP_PASSWORD}

WhatsApp (via Baileys):

enabled: true
session_path: ~/.hermes/whatsapp-session
qr_code_callback: true

Chat Interface Usage

Basic Chat

Type naturally or use slash commands:

> What's the weather in San Francisco?
[Hermes uses web search tool, returns answer]

> /web latest news on AI regulation
[Forces web search, streams results]

> /image a cyberpunk city at sunset
[Generates image via FAL.ai or configured provider]

Slash Commands

Token Usage Tracking

Bottom of chat shows live counts:

📊 Prompt: 1,234 tokens • Completion: 567 tokens • Cost: $0.0123

Use /usage for detailed breakdown:

> /usage
Session Usage:
  Total Prompt Tokens: 12,345
  Total Completion Tokens: 5,678
  Total Cost: $0.123
  Messages: 15

Skills Management

Installing Skills

Skills screen:

1. Browse bundled skills (pre-installed with Hermes)
2. Or install from GitHub:

   Repository: username/repo-name
   Branch: main (optional)
   

3. Click Install

Skill Structure

Skills are markdown files with YAML frontmatter:

---
name: web-search-expert
description: Expert at web searching and information retrieval
triggers:
  - "search the web for"
  - "find information about"
  - "look up"
---

# Web Search Expert

Use the web_search tool to find current information...

## Examples

When user asks: "What's the latest on GPT-5?"
1. Use web_search with query "GPT-5 latest news"
2. Summarize findings
3. Cite sources

Editing Skills

Skills screen → Click skill → Edit:

• Modify triggers, description, or content
• Changes saved to ~/.hermes/skills/<name>/SKILL.md

Memory System

Viewing Memory

Memory screen:

• Memory Entries: List of stored facts/context
• User Profile: Persistent user information
• Capacity: Current usage vs. limit

Adding Memory

Chat naturally — Hermes auto-saves important context:

> My name is Alice and I prefer Python over JavaScript
[Hermes stores to memory automatically]

> What's my name?
[Hermes retrieves: "Alice"]

Or explicit:

> Remember I'm working on a TypeScript project called Hermes Desktop
[Hermes: ✓ Stored to memory]

Editing Memory Entries

Memory screen:

1. Click entry to view
2. Edit content or delete
3. Save changes

Memory Providers

Configured in Memory screen → Configure Providers:

• Honcho: Managed memory service
• Hindsight: Self-hosted memory
• Mem0: Personalized AI memory
• RetainDB: Vector memory DB
• Supermemory: Context-aware memory
• ByteRover: Local memory store

Session Management

Searching Sessions

Sessions screen:

• Search bar: Full-text search (SQLite FTS5) across all conversations
• Date groups: Conversations grouped by Today, Yesterday, Last 7 Days, Last 30 Days, Older

// Example: Search sessions programmatically (if extending app)
import { searchSessions } from './main/database';

const results = await searchSessions('typescript error handling');
// Returns: [{ id, title, timestamp, snippet, profileId }, ...]

Resuming Sessions

1. Sessions screen → Click conversation
2. Chat loads with full history
3. Continue conversation from last message

Deleting Sessions

Right-click session → Delete or click trash icon.

Tools Management

Tools screen → Enable/disable toolsets:

Example enabling web tool:

# ~/.hermes/profiles/default/config.yaml
tools:
  web:
    enabled: true
    exa_api_key: ${EXA_API_KEY}
    tavily_api_key: ${TAVILY_API_KEY}

Scheduled Tasks

Schedules screen → Create Task:

name: "Daily standup summary"
schedule: "0 9 * * 1-5"  # 9 AM weekdays
task: "Summarize yesterday's GitHub activity and send to Slack"
delivery:
  type: slack
  channel: "#standup"

Schedule types:

• Minutes: Every N minutes
• Hourly: Every N hours
• Daily: Specific time daily
• Weekly: Specific day/time weekly
• Custom: Full cron expression

Delivery targets: Telegram, Discord, Slack, WhatsApp, Signal, Matrix, Email, SMS, Webhook, Home Assistant, File, Terminal, Memory, Chat (in-app), Desktop Notification

Development Setup

Prerequisites

node --version  # v18+ recommended
npm --version   # v9+

Clone and Install

git clone https://github.com/fathah/hermes-desktop.git
cd hermes-desktop
npm install

Development Mode

npm run dev

This starts:

• Vite dev server for React UI
• Electron main process with hot reload
• TypeScript watch compiler

Project Structure

hermes-desktop/
├── src/
│   ├── main/          # Electron main process
│   │   ├── index.ts   # Entry point, IPC handlers
│   │   ├── database.ts # SQLite sessions/memory
│   │   ├── installer.ts # Hermes install logic
│   │   └── updater.ts  # Auto-update
│   ├── preload/       # Electron preload script
│   │   └── index.ts   # IPC bridge to renderer
│   └── renderer/      # React UI
│       ├── App.tsx
│       ├── screens/   # Chat, Sessions, Tools, etc.
│       ├── components/
│       └── lib/       # SSE parser, utils
├── electron.vite.config.ts
├── package.json
└── resources/         # Icons, installers

IPC Communication Pattern

Renderer → Main:

// src/renderer/lib/api.ts
export async function sendChatMessage(message: string, profileId: string) {
  return window.electron.ipcRenderer.invoke('chat:send', { message, profileId });
}

Main handler:

// src/main/index.ts
ipcMain.handle('chat:send', async (event, { message, profileId }) => {
  const response = await fetch('http://127.0.0.1:8642/chat', {
    method: 'POST',
    headers: { 'Content-Type': 'application/json' },
    body: JSON.stringify({ message, profile: profileId })
  });
  // Stream SSE events back to renderer
});

SSE Streaming Implementation

// src/renderer/lib/sse-parser.ts
export class SSEParser {
  private buffer = '';

  parse(chunk: string): SSEEvent[] {
    this.buffer += chunk;
    const lines = this.buffer.split('\n');
    this.buffer = lines.pop() || '';
    
    const events: SSEEvent[] = [];
    let currentEvent: Partial<SSEEvent> = {};
    
    for (const line of lines) {
      if (line.startsWith('event:')) {
        currentEvent.event = line.slice(7).trim();
      } else if (line.startsWith('data:')) {
        currentEvent.data = line.slice(6).trim();
      } else if (line === '') {
        if (currentEvent.event) {
          events.push(currentEvent as SSEEvent);
        }
        currentEvent = {};
      }
    }
    
    return events;
  }
}

Usage in Chat:

// src/renderer/screens/Chat.tsx
const parser = new SSEParser();

fetch('http://127.0.0.1:8642/chat', {
  method: 'POST',
  body: JSON.stringify({ message })
}).then(async (response) => {
  const reader = response.body!.getReader();
  const decoder = new TextDecoder();
  
  while (true) {
    const { done, value } = await reader.read();
    if (done) break;
    
    const chunk = decoder.decode(value);
    const events = parser.parse(chunk);
    
    for (const event of events) {
      if (event.event === 'content') {
        appendContent(JSON.parse(event.data).text);
      } else if (event.event === 'tool_start') {
        showToolProgress(JSON.parse(event.data));
      } else if (event.event === 'usage') {
        updateTokenCount(JSON.parse(event.data));
      }
    }
  }
});

Building for Production

Build All Platforms (current OS)

npm run build

Output: dist/ directory with installers

Platform-Specific Builds

npm run build:win    # Windows .exe
npm run build:mac    # macOS .dmg
npm run build:linux  # .AppImage, .deb, .rpm

Build Configuration

Edit electron-builder.yml:

appId: com.nousresearch.hermesdesktop
productName: Hermes Agent
directories:
  output: dist
  buildResources: resources

mac:
  target:
    - dmg
    - zip
  category: public.app-category.productivity
  icon: resources/icon.icns

win:
  target:
    - nsis
  icon: resources/icon.ico

linux:
  target:
    - AppImage
    - deb
    - rpm
  category: Utility
  icon: resources/icon.png

Troubleshooting

Installer Stalls on "Installing dependencies"

WSL: Grant temporary passwordless sudo (see Installation section)

General: Check logs in ~/.hermes/logs/install.log:

tail -f ~/.hermes/logs/install.log

"Connection refused" on 127.0.0.1:8642

Hermes not running. Start manually:

cd ~/.hermes
source venv/bin/activate
python -m hermes.server

Or restart from Settings → Restart Hermes.

Remote mode connection fails

1. Verify URL is reachable: curl https://your-hermes-server.com/health
2. Check API key in remote Hermes: hermes keys list
3. Ensure firewall allows traffic

Memory provider not connecting

Check API keys in Memory → Configure Providers:

// Test Honcho connection
const response = await fetch('https://api.honcho.dev/apps', {
  headers: { 'Authorization': `Bearer ${process.env.HONCHO_API_KEY}` }
});

Gateway fails to send messages

1. View logs: Settings → View Logs → Select gateway
2. Telegram: Verify bot token with @BotFather, check allowed_users IDs
3. Discord: Ensure bot has SEND_MESSAGES permission in channel
4. Email: Test SMTP credentials with telnet smtp.server.com 587

Skills not triggering

1. Check triggers in skill YAML frontmatter are specific
2. Ensure skill is enabled: Skills screen → Check toggle
3. Restart profile: Agents → Switch away and back

Auto-update fails (Fedora RPM)

RPM builds don't support auto-update. Download new .rpm and reinstall:

sudo dnf install ./hermes-desktop-<new-version>.rpm --nogpgcheck

Chat stuck on "Thinking..."

1. Check network: curl -I http://127.0.0.1:8642/health
2. Check model provider API status (OpenRouter, Anthropic, etc.)
3. View debug logs: /debug in chat or Settings → View Logs → Agent

Session search returns no results

Database index corrupt. Rebuild FTS5 index:

cd ~/.hermes/profiles/default
sqlite3 sessions.db "DELETE FROM sessions_fts; INSERT INTO sessions_fts SELECT * FROM sessions;"

macOS "App is damaged" error

Remove quarantine:

xattr -cr "/Applications/Hermes Agent.app"

Windows SmartScreen blocks installer

Click "More info" → "Run anyway" (app is not code-signed).

Environment Variables

# LLM Providers
OPENROUTER_API_KEY=sk-or-...
ANTHROPIC_API_KEY=sk-ant-...
OPENAI_API_KEY=sk-...
GOOGLE_API_KEY=...
XAI_API_KEY=...

# Search/Web Tools
EXA_API_KEY=...
TAVILY_API_KEY=...
FIRECRAWL_API_KEY=...

# Image Generation
FAL_API_KEY=...
REPLICATE_API_TOKEN=...

# Memory Providers
HONCHO_API_KEY=...
MEM0_API_KEY=...

# Messaging Gateways
TELEGRAM_BOT_TOKEN=...
DISCORD_BOT_TOKEN=...
SLACK_BOT_TOKEN=...
TWILIO_ACCOUNT_SID=...
TWILIO_AUTH_TOKEN=...

# Email
EMAIL_USERNAME=...
EMAIL_APP_PASSWORD=...

# Analytics
WANDB_API_KEY=...

Store in ~/.hermes/.env or OS-level environment.

Testing

Run Tests

npm test

Test suites:

• SSE parser: Validates event parsing, incomplete chunks, malformed data
• IPC handlers: Mock Electron IPC, test chat/session/profile handlers
• Preload API: Ensures all main process APIs are safely exposed
• Installer utils: Dependency checking, path resolution
• Constants: Validates config schema, provider definitions

Example Test

// src/__tests__/sse-parser.test.ts
import { describe, it, expect } from 'vitest';
import { SSEParser } from '../renderer/lib/sse-parser';

describe('SSEParser', () => {
  it('parses complete events', () => {
    const parser = new SSEParser();
    const events = parser.parse('event: content\ndata: {"text":"Hello"}\n\n');
    
    expect(events).toHaveLength(1);
    expect(events[0].event).toBe('content');
    expect(JSON.parse(events[0].data).text).toBe('Hello');
  });
  
  it('buffers incomplete events', () => {
    const parser = new SSEParser();
    const events1 = parser.parse('event: content\n');
    const events2 = parser.parse('data: {"text":"Hi"}\n\n');
    
    expect(events1).toHaveLength(0);
    expect(events2).toHaveLength(1);
  });
});

Resources

• Documentation: https://hermes-agent.nousresearch.com/docs/
• GitHub Repository: https://github.com/fathah/hermes-desktop
• Releases: https://github.com/fathah/hermes-desktop/releases
• Telegram Community: https://t.me/hermes_agent_desktop
• Issue Tracker: https://github.com/fathah/hermes-desktop/issues
• Hermes Agent Core: https://github.com/NousResearch/hermes-agent

License

MIT License — see LICENSE