# hermes-agent **Repository Path**: cndavy/hermes-agent ## Basic Information - **Project Name**: hermes-agent - **Description**: No description available - **Primary Language**: Unknown - **License**: MIT - **Default Branch**: main - **Homepage**: None - **GVP Project**: No ## Statistics - **Stars**: 0 - **Forks**: 0 - **Created**: 2026-04-13 - **Last Updated**: 2026-04-13 ## Categories & Tags **Categories**: Uncategorized **Tags**: None ## README # 🪽 Hermes Agent: Lightweight Installer [![Platform](https://img.shields.io/badge/platform-Windows%20%7C%20macOS-lightgrey)](#-installation) [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
Hermes Agent: Lightweight Installer.
The official lightweight installer for **Hermes Agent** — an autonomous, self-improving AI agent by [Nous Research](https://github.com/NousResearch/hermes-agent). This installer allows you to deploy the agent locally in a few clicks, without the need to manually configure Python environments (`uv`, `venv`), databases for memory, or mess with system dependencies. --- ## 🎯 What is Hermes Agent and why do you need it? **Hermes Agent** is an AI that grows with you. It is not just code autocomplete in your IDE or a chatbot wrapper around an API. It is a full-fledged digital employee that runs continuously on your hardware (or VPS), remembers everything it learns, and becomes more independent with every session. **Why do you need it?** * **Unattended Automation:** You can assign it to monitor servers, parse data, test code, or write reports. The agent will run in the background 24/7. * **Escaping "Groundhog Day":** You no longer need to feed the project context to the AI every single time. Hermes reads its own past logs and already knows who you are, what your tech stack is, and where you left off yesterday. * **Delegating Dev/Ops Tasks:** The agent can write code on its own, test it in an isolated sandbox, fix errors, and deploy the result while you drink your coffee. --- ## ✨ Key Features (Core Functionality) Our installer deploys the full Hermes core with all its official features: * 🧠 **Self-Improving Memory Loop:** The agent uses full-text search (SQLite FTS5) across past sessions and integrates with `Honcho` for deep user profiling. When completing complex tasks, it **automatically creates new skills** (compatible with the `agentskills.io` standard) to solve the problem faster next time. * 🌐 **Multi-platform Gateway:** One mind — multiple interfaces. Start a task in the terminal (CLI), and receive answers and manage the agent from your smartphone via **Telegram, Discord, Slack, WhatsApp, or Signal**. * 🎙️ **Voice Mode:** Supports Push-to-Talk in the terminal and voice messages in messengers. Transcription is done locally and privately using the built-in `faster-whisper`. * 🛡️ **Real Sandboxing:** 6 backends for safely executing generated code: Local, Docker, SSH, Daytona, Singularity, and Modal. Your core system files are always protected from accidental AI actions. * ⏱️ **Natural Language Scheduled Automations:** The built-in cron allows you to schedule tasks in plain text. Write: *"Every day at 9:00 AM, check the error logs on prod and send me a summary in Telegram"*, and the agent will set everything up autonomously. * ⚡ **Delegation and Parallelization:** The agent can spawn isolated sub-agents for parallel work or write Python scripts that call its own tools via RPC (reducing multi-step processes to zero context cost). * 🔌 **First-Class Plugins and Model Agnosticism:** Extend the agent by simply dropping Python files into the `~/.hermes/plugins/` folder. Switch on the fly between OpenRouter (200+ models), native Anthropic API (with prompt caching), Vercel AI Gateway, or local LLMs. --- ## 🌟 Why the Lightweight Installer? The core of Hermes Agent requires complex system orchestration. Our installer handles all the heavy lifting: * **Zero-Dependencies:** Does not require pre-installed Python 3.11+. All runtime components are already packaged inside. * **Auto-Sandboxing:** Automatically detects Docker on your PC (or prompts you to install it) and configures containers for the agent. * **Smart Updates:** A built-in background manager automatically downloads new Hermes core releases from Nous Research while preserving your memory database. --- ## 🚀 Installation Choose the appropriate package for your operating system on the [Releases](../../releases) page. ### 🪟 Windows (Windows 10 / 11) 1. Download the latest version of `HermesAgent_x64.exe`. 2. Run the file and follow the setup wizard instructions. 3. The installer will automatically add the `hermes` CLI utility to your system `PATH`. ### 🍎 macOS (Apple Silicon) 1. Download `HermesAgent_macOS.dmg`. 2. Double-click the `.dmg` file to mount it. 3. Drag and drop the **Hermes Agent** icon into your `Applications` folder. 4. Upon first launch, go to *System Settings -> Privacy & Security* and click "Open Anyway". --- ## ⚙️ First Launch and Configuration Open your terminal and run the initialization command. An interactive wizard will launch: ```bash hermes init ``` You will be asked to configure three basic things: 1. **LLM Provider:** Choose OpenRouter, Anthropic, or local inference. 2. **API Key:** Paste your key. 3. **Execution Backend:** Choose `Docker` for secure isolation of executed code. ### Starting the Core Run the agent in the background: ```bash hermes start ``` Check the status of the modules (Memory, Scheduler, Sandbox): ```bash hermes status ``` --- ## 💬 Connecting to Messengers (Messaging Gateway) Hermes can live on your server or laptop while communicating with you via a messenger. For example, to link Telegram: 1. Create a bot via `@BotFather` in Telegram and copy the token. 2. Run the following command in your terminal: ```bash hermes gateway add telegram --token YOUR_BOT_TOKEN ``` 3. Start the gateway: ```bash hermes gateway start ``` You're all set! Now your context in the console and on your phone is synchronized. Start messaging the bot, and the agent will reply. --- ## 🗄 Directory Structure The installer creates a `~/.hermes` directory (or `%USERPROFILE%\.hermes` on Windows) where your agent's mind lives: * `/config.json` — Core settings and API key routing. * `/memory/` — SQLite databases containing session context and user profiles. * `/skills/` — Automatically generated and downloaded skills (`agentskills.io` format). * `/plugins/` — Directory for custom tools. * `/logs/` — Structured daily logs (the foundation for the AI's long-term memory). --- ## ❓ FAQ (Frequently Asked Questions) ### 1. Do I have to install Docker? **No, but we highly recommend it.** By default, the installer will try to use Docker to create an isolated sandbox. If Docker is not installed, you can choose the `local` backend during initialization (`hermes init`). In this case, the agent will execute code directly in your OS with your user's privileges, which requires caution. ### 2. Can Hermes Agent run 100% offline? **Yes.** The installer contains all necessary local dependencies (including `faster-whisper` for audio transcription and the database). To detach from cloud APIs, simply install [Ollama](https://ollama.com/), download any model (e.g., `ollama run hermes-3-llama-3.1-8b`), and select the local provider during `hermes init`. Your data will never leave your device. ### 3. How does the agent avoid overflowing the LLM context window if it remembers "everything"? Hermes does not send your entire chat history to the model. It uses a hybrid memory system: lightweight full-text search (SQLite FTS5) and background summarization. When you ask a question, the agent first makes a hidden query to its database, extracts only relevant facts about you or the project, and adds only those to the current prompt. ### 4. How do I transfer the agent's "brain" to another computer or server? The agent's entire personality, generated skills, settings, and memory base are stored in a single folder: `~/.hermes` (or `%USERPROFILE%\.hermes` on Windows). To move the agent to another PC or VPS, simply install our Lightweight Installer there and copy this folder over, replacing the existing one. The agent will "wake up" in the new location without losing any context. ### 5. Is it safe to give the agent access to Telegram or Discord? **Absolutely.** Messengers are connected via the built-in Messaging Gateway. Only you have access to the bot (the gateway filters messages by your User ID). Furthermore, if the agent decides to execute a complex script via Telegram, it will still run it inside an isolated Docker container on your server, not in your main host system. ---