Skill Level: Advanced Type: Browser Extension / Agent Tool Pricing: Free & Open Source Status: Trending on GitHub
If you use Hermes Agent by Nous Research, you already know the power of local, autonomous AI workflows. But copying and pasting URLs, DOM snippets, and text into your terminal is a massive friction point. Built by community creator Jon Komet (@abundantbeing), the Hermes Browser Extension bridges this gap. It acts as a native browser side panel that securely pipes your active web context directly into your local or remote Hermes Agent runtime—giving your AI eyes and hands on the web without compromising your privacy.
Want to inspect the source code or follow the project roadmap? Check out the official repository here: 🔗 github.com/abundantbeing/hermes-browser-extension
What is the Hermes Browser Extension?
The Hermes Browser Extension is a browser-native conduit designed to connect whatever is inside your active browser tab to a running instance of Hermes Agent. Crucially, the extension itself does not contain any bulky Large Language Models (LLMs). Instead, it acts as an intelligent client that communicates via WebSocket and REST APIs to your local Hermes Gateway (typically running on port 8642) or a remote Hermes server.
Because it hooks directly into your Hermes runtime, it inherits all your existing configurations automatically—your models, custom system prompts, skills, memory providers, and Model Context Protocol (MCP) tool servers remain fully intact. It introduces a dedicated Side Panel UI where you can chat with Hermes, execute slash commands against live DOM elements, isolate session context per tab, and utilize dual-engine voice dictation without relying on third-party cloud wrappers.
Who is it for?
- AI Agents & Automation Engineers: Developers building autonomous workflows who need an easy way to feed live, authenticated browser states (like internal dashboards or gated web apps) into a local AI agent.
- Privacy-Conscious Researchers: Analysts and readers who want AI-powered web summarization, cross-tab comparison, and Q&A using local weights (like Llama 3 or Qwen) rather than sending their browsing history to OpenAI or Google.
- Power Users & Developers: Engineers already utilizing Hermes Agent in the terminal who want a GUI companion right inside Chrome, Edge, or Brave to execute quick refactoring or explanation commands on web-hosted code repositories and documentation.
What makes it special?
- Transparent “What Hermes Saw” Receipt: After every turn, the extension generates a collapsible audit receipt showing the exact markdown, DOM hierarchy, and metadata extracted from your webpage. You never have to guess what context was sent to your model.
- Granular Context Scope Control: You can dynamically switch between attaching the full active tab, pinning a specific tab, selecting multiple open tabs, feeding only highlighted text, or disabling web context entirely for pure chat.
- Untrusted Context Wrapping: To defend against hidden prompt injection attacks embedded in malicious web pages, all extracted web text is automatically wrapped and labeled as untrusted context before hitting your LLM parser.
- Smart Dual-Engine Voice Dictation: Features a built-in microphone integration that opens a dedicated, secure voice tab to capture speech without running into browser side-panel permission restrictions, piping transcripts straight back into the composer.
- Zero Global Mutation: Any model switches, parameter tweaks, or session bindings you apply inside the extension stay strictly isolated to your browser sessions; they will never overwrite your global terminal defaults in
~/.hermes/config.yaml.
Requirements before you start
Because this is a bridge tool, you need a functioning backend runtime before installing the browser client:
- Hermes Agent Runtime: You must have a working installation of Hermes Agent running locally or accessible on a remote server.
- Node.js & npm: Node.js version 20.0 or higher is required to compile the extension asset bundle from source.
- Supported Web Browser: Google Chrome (version 114+ baseline), Microsoft Edge, Brave, Comet, or any modern Chromium-based browser with native Side Panel API support.
- Network / Gateway Access: Your local Hermes Gateway/API server must be enabled and listening (defaults to
http://127.0.0.1:8642).
Step-by-step installation
Step 1 — Clone the Repository and Build from Source
Open your terminal and clone the GitHub repository. Once downloaded, navigate into the project directory, install the Node dependencies, and run the production build script. This will compile the TypeScript and UI components into a loadable Chromium extension inside a new dist/ folder:
# Clone the repo and build the extension
git clone https://github.com/abundantbeing/hermes-browser-extension.git
cd hermes-browser-extension
npm install
npm run buildStep 2 — Load the Unpacked Extension in Your Browser
Because this is a developer-focused tool, you will install it directly in developer mode rather than from an extension store:
# 1. Open your Chromium browser and navigate to:
chrome://extensions # (or edge://extensions for Microsoft Edge)
# 2. Toggle the "Developer mode" switch in the top right corner.
# 3. Click the "Load unpacked" button in the top left.
# 4. Select the generated "dist" folder inside your cloned repo directory!
# (IMPORTANT: Select the /dist folder, NOT the repo root!)Step 3 — Launch and Connect to Your Hermes Gateway
Pin the Hermes icon to your browser toolbar and click it (or press Alt+H) to open the side panel. Now, let’s link it to your backend:
# Ensure your Hermes Agent API gateway is running in your terminal:
hermes gateway --port 8642
# In the extension side panel:
# 1. Click "Manual setup" (or "Connect to Hermes" if using Desktop approval).
# 2. Select "Local gateway".
# 3. Set the Gateway URL to: http://127.0.0.1:8642
# 4. Paste your API_SERVER_KEY (found in ~/.hermes/.env or config).
# 5. Click "Test connection" -> If green, click "Save settings"!Step 4 — Test Web Context Extraction
Navigate to any article or GitHub repository in your browser. In the Hermes side panel composer, type a quick slash command like /summarize or ask: “What are the main dependencies listed on this page?” Expand the What Hermes saw receipt below your message to verify that the extension successfully scraped and formatted the DOM for your local agent!
Common errors and fixes
| Error | Meaning | Fix |
|---|---|---|
Failed to load extension: Manifest file is missing | You selected the root project directory instead of the compiled build folder when clicking “Load unpacked”. | Ensure you ran npm run build successfully, then select the generated dist/ subfolder inside the repository directory. |
Connection Refused (http://127.0.0.1:8642) | The extension cannot reach your local Hermes Agent backend server. | Verify that Hermes Agent is actively running on your host machine and that the API server module is listening on port 8642. Check for firewalls blocking loopback connections. |
401 Unauthorized / Invalid API Key | The security token passed to the extension does not match the key expected by the Hermes Gateway. | Open your terminal and check your ~/.hermes/.env file for your actual API_SERVER_KEY or scoped browser token. Copy and paste it cleanly into the extension settings. |
Microphone access denied in Side Panel | Chromium security policies restrict direct hardware media capture inside iframe-like extension side panels. | Use the built-in voice workflow: click the mic icon, allow the extension to open the companion Voice Dictation Tab, and click “Start dictation” there to satisfy Chromium’s gesture requirement. |
DOM Scraping Failed / Empty Context | You are trying to extract context from a restricted system browser page (e.g., chrome://settings or the Chrome Web Store). | Browser security model forbids content scripts from injecting into internal chrome:// or extension store URLs. Navigate to a standard HTTP/HTTPS webpage and try again. |
Free vs Paid comparison
| Feature / Metric | Hermes Extension + Local Agent (Free) | Commercial AI Browser Assistants (Paid Cloud) |
|---|---|---|
| Data Sovereignty | 100% Private. Scraped page DOMs, tokens, and cookies remain on your machine or private gateway. | Page context and browsing behaviors are transmitted to commercial cloud endpoints for processing. |
| Tool & Skill Ecosystem | Full access to your local Hermes skills, MCP servers, Langfuse tracing, and local memory databases. | Restricted to the proprietary web search and plugins built by the commercial vendor. |
| Customization & Scope | Deep granular control over DOM extraction, prompt wrapping, and per-tab session isolation. | Usually limited to simple “Summarize Page” or generic page chat buttons. |
| Cost | $0 / Free forever under open-source licensing. Requires your own local/remote compute resources. | Typically requires a $20/month subscription (e.g., ChatGPT Plus, Claude Pro, or Copilot Pro). |
Bottom line: The Hermes Browser Extension transforms Hermes Agent from a brilliant terminal tool into a true, everyday web companion. By securely piping your active browser context into your local runtime without third-party cloud telemetry, it delivers the holy grail of browser automation: total data privacy, deep MCP tool integration, and frictionless web AI.
Alternatives — 3 similar tools
1. Open WebUI (with Chrome Extension)
Open WebUI is a highly popular, self-hosted web interface designed to interact with local LLMs (like Ollama or vLLM). It features an official companion browser extension that allows you to clip webpage content and send it directly to your self-hosted chat interface. While fantastic as a general-purpose ChatGPT replacement, it lacks the deep autonomous agent capabilities, local skill execution, and MCP tool integrations native to Hermes.
🔗 openwebui.com
2. Harpa AI
Harpa AI is a powerhouse browser extension that combines AI chat, web automation, price tracking, and SERP enrichment into a single side panel. It features a massive library of pre-built marketing and copywriting prompts and can connect to local LLMs or cloud providers. However, it is a proprietary tool with heavy UI bundling, making the lightweight, developer-first, open-source approach of the Hermes extension much more appealing for engineers.
🔗 harpa.ai
3. Page Assist
Page Assist is an open-source browser extension designed as a simple visual sidebar for local AI models running via Ollama or LM Studio. It allows you to chat with your local models and interact with the text of your active web page. It is much easier to set up for casual users who only want page summarization, but it cannot match the advanced agentic workflows, multi-tab orchestration, and execution capabilities of a full Hermes Agent runtime.
🔗 github.com/nilsherzig/LLocalSearch
🚀 Want more free AI tools like this?
We find, test, and write setup guides for the best free and open-source AI tools — so you don’t have to dig through GitHub yourself.Browse Free AI Tools at globalaiforce.com/shop →
📸 Follow us for daily AI tool tips and tutorials: instagram.com/globalaiforce