Published: 2026-06-03

Hermes Desktop + Ollama: Install the GUI and Wire In a Local Model

Hermes Desktop is a new official GUI that replaces terminal-based interaction with a full chat interface, model picker, MCP tool manager, and persona system. Fahd Mirza demos the complete setup on Ubuntu: updating an existing Hermes install, launching the desktop for the first time, connecting an Ollama-hosted local model, and exploring themes, skills, messaging integrations (Telegram, Discord), and session management.

Source video

"Hermes Desktop + Ollama: Run a Self-Improving AI Agent on Your Own Server" by Fahd Mirza — Watch on YouTube →

Key Takeaways

hermes update brings an existing install up to date — run this before launching the desktop for the first time to ensure all Node.js packages are current.
hermes desktop launches the GUI; the first run installs dependencies and may take a minute or two. Works on Ubuntu, Windows, and macOS.
Model providers available in the GUI: API-based (Gemini, GLM, MiniMax, and others), local Ollama endpoints, custom OpenAI-compatible endpoints, and hosted gateways.
MCP servers can be toggled on/off individually in the GUI — no config file editing required. Messaging channels (Telegram, Discord) have dedicated skill cards with token/ID input fields.
Multiple personas (profiles) are supported, and sessions can be exported, pinned, archived, or deleted from the sidebar.

Commands & Code Mentioned

# Update Hermes to latest (run before first desktop launch)
hermes update

# Launch Hermes Desktop GUI
hermes desktop

# Connect Ollama model: configure in GUI under Settings → Local Providers
# Select Ollama, paste your endpoint (default: http://localhost:11434)
# Choose model from the dropdown (e.g. qwen3:6b, llama3.2, etc.)

# Alternative: set model via CLI before launching
hermes set model ollama/qwen3:6b

Desktop GUI Features at a Glance

Themes: Default, Dark, Midnight, Amber, Cyberpunk — set under Settings → Appearance.
Multi-model chat: Switch between any configured provider mid-session from the model picker in the chat header.
Skills & tools: Browse by category (General, Creative, Data Science, Messaging); toggle MCPs on/off per skill.
Session management: Pin, rename, export, archive, or delete sessions from the sidebar.
Memory: Built-in memory settings plus external memory configuration (Obsidian and others).
Advanced parameters: Inference hyperparameters, refresh rates, exponential backoff retries — all configurable without editing files.

Related Hermes Guides

Hermes Setup Guide — full installation from scratch
Hermes Web Dashboard — the localhost:9119 control panel
Hermes Memory & Persistence — external memory via Obsidian and other tools
Hermes MCP Tools — connecting and managing MCP integrations