Use Hermes Agent

Hermes Agent is an open source autonomous agent built by Nous Research . It runs autonomously for long stretches, builds persistent memory, spawns subagents in parallel, and reaches you on Telegram, Discord, Slack, and other platforms while working on a remote VM. From version v0.13.0, Hermes has native support for Lightpanda as a browser backend.

When configured to use Lightpanda, Hermes will route browser actions through Lightpanda, with automatic Chrome fallback for actions Lightpanda doesn’t yet support (such as screenshots).

Prerequisites

Hermes Agent v0.13.0 or later
Lightpanda installed and on your PATH

If you don’t yet have Lightpanda installed, follow the installation guide.

Install Hermes

If you don’t have Hermes Agent yet, install it with the official one-liner:


curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash

If you already have Hermes installed, update to v0.13.0 or later:


hermes update

Confirm the version:


hermes --version

Configure Lightpanda as the browser backend

Open Hermes’ config file:


open ~/.hermes/config.yaml

Find the browser: block and add the engine: lightpanda line:


browser:
  engine: lightpanda

Save and close the file.

Verify the configuration

Start Hermes:


hermes

Run the /browser slash command:


/browser

You should see:


🌐 Browser: local Lightpanda (agent-browser --engine lightpanda)
   ⚡ Lightpanda: faster navigation, no screenshot support
   Automatic Chrome fallback for screenshots and failed commands

Hermes will now route browser actions through Lightpanda.

Run a quick test

Inside Hermes, try a prompt that involves browsing:


Visit https://news.ycombinator.com and list the top 5 story titles.

Hermes will spawn its browser tools and route them through Lightpanda. You’ll see streaming tool calls (browser_navigate, browser_snapshot) executing against Lightpanda.

Automatic Chrome fallback

Lightpanda doesn’t yet support every browser action Hermes might call. When Hermes invokes an unsupported action (e.g. screenshots), or when a command errors out on Lightpanda, Hermes transparently retries on Chrome.

This means enabling engine: lightpanda is non-disruptive: Lightpanda handles the actions it supports, Chrome handles the rest, and your agent’s behavior remains unchanged.

The current set of actions supported by Lightpanda inside Hermes covers the core agent workflow: navigate, snapshot, click, type, scroll, back, press, and eval. Actions that fall back to Chrome include screenshots, PDF generation, file uploads, and clipboard operations.