Browser logo

Browser

by lightpanda-io

Headless browser with a built-in MCP server for AI-driven navigation, page interaction, and web data extraction.

Install Browser

Run the project’s own setup commands, in order.

Setup steps from the project

lightpanda-io/browser installs with its own commands rather than a config we can generate, so these are copied as-is. Run them in order in a terminal. Source: github.com

  1. 1. Install
    brew install lightpanda-io/browser/lightpanda
    
  2. 2. Install
    curl -L -o lightpanda https://github.com/lightpanda-io/browser/releases/download/nightly/lightpanda-x86_64-linux && \
    chmod a+x ./lightpanda
    

Overview

lightpanda-io/browser provides a headless browser built from scratch in Zig, with an MCP server for AI agents and web automation. It can navigate pages, click through flows, fill forms, and extract structured data, while also supporting CDP and WebDriver BiDi connections. Reach for it when browser workloads need a lightweight runtime, direct agent control, or command-line page dumps.

Use cases

  • Fetch pages as HTML, Markdown, images, PDFs, or structured metadata
  • Drive browser workflows with natural-language agent tasks
  • Replay saved browser interactions as deterministic PandaScripts
  • Connect Puppeteer automation through the CDP endpoint
  • Run lightweight crawling workloads on Linux or macOS

Key features

  • Built-in MCP server for AI agents
  • Headless browser implemented in Zig
  • CDP and WebDriver BiDi server modes
  • HTML, Markdown, PNG, PDF, and structured-data dumps
  • Agent mode with PandaScript export and replay
  • Official Docker images for Linux amd64 and arm64

Security & health

What it can reach and how reliably it runs — from our own checks, not the maintainer’s claims.

ActiveLast checked Today

Recent health check succeeded.

Adoption & maintenance

Factual signals from GitHub, npm, and our automated checks — not a rating.

GitHub stars41kStargazers on the source repository.
Last commitTodayMost recent push to the default branch.
Directory activity9 viewsConfig copies, upvotes, and views on AllMCPs.

How Browser works

What lightpanda-io/browser does

lightpanda-io/browser combines a headless browser with a built-in Model Context Protocol server for AI agents and automation. Lightpanda is implemented in Zig rather than being a Chromium or WebKit variant. Its browser-facing capabilities include navigating to pages, interacting with links and forms, and extracting structured content.

The command-line interface can fetch a URL and emit HTML, Markdown, PNG, PDF, or structured metadata. Structured-data output includes JSON-LD, OpenGraph fields, Twitter card data, meta tags, and link tags. The browser also has an agent mode that accepts plain-language tasks, drives the browser through interactions, and can save the resulting workflow as a PandaScript for later replay.

A published crawler benchmark requested 933 real web pages over the network on an AWS EC2 m5.large instance. In that test, Lightpanda used 123 MB of peak memory for 100 pages and completed in 5 seconds, compared with 2 GB and 46 seconds for headless Chrome.

How it works

Run the binary directly for one-off page retrieval, or start its server mode to expose a browser endpoint. The server can listen on TCP port 9222 or on a Unix-domain socket. Automation code can connect through Puppeteer using the browser's WebSocket endpoint. WebDriver BiDi is available by starting the server with the WebDriver protocol, and CDP and BiDi can be enabled together.

The agent mode runs in the same process as the browser. It can navigate sites, click through workflows, fill forms, and extract structured data. A saved session becomes JavaScript containing Lightpanda browser primitives; it can be replayed with the run command without requiring a model at runtime. Agent sessions can use supported hosted or local model providers, or run with --no-llm for a direct REPL.

Page dumps support waiting controls such as a delay, a selector, a script, or a wait-until condition. The --obey-robots option is available for fetch and server commands.

Setup and configuration

Install the latest nightly build with Homebrew or the Arch Linux User Repository, download a Linux or macOS binary, use the Nix flake, or run the official Docker image. Docker images are available for Linux amd64 and arm64. A container can expose the CDP endpoint on 127.0.0.1:9222.

Linux binaries require glibc. The documented x86_64 builds need glibc 2.35 and arm64 builds need glibc 2.38; Alpine and other musl-based distributions do not run them directly. Native Windows binaries are not provided, but the browser can run inside WSL2. Android and Termux are not supported as native environments.

After downloading a binary, verify it with ./lightpanda version. Start a browser endpoint with lightpanda serve, fetch a page with lightpanda fetch, or launch the interactive agent with lightpanda agent. The repository also provides a source-build path for environments that cannot use the release binaries.

Reviews

No reviews yet — be the first to share how it worked for you.

Frequently Asked Questions about Browser

lightpanda-io/browser is an MCP server that connects AI agents to the Lightpanda headless browser and the web pages it loads. Its browser capabilities cover navigation, clicking, form filling, and structured-data extraction, with additional access through command-line fetches, CDP, WebDriver BiDi, and agent-driven workflows.

Install Lightpanda with Homebrew, the Arch Linux User Repository, a nightly Linux or macOS binary, Nix, or the official Docker image. Verify a downloaded binary with `./lightpanda version`; Docker users can expose the browser endpoint on port 9222.

Yes, Puppeteer can connect through Lightpanda's CDP WebSocket endpoint. Start `lightpanda serve`, then pass its `ws://127.0.0.1:9222` address as Puppeteer's `browserWSEndpoint`.

It supports Windows through WSL2 rather than a native Windows binary. Install Lightpanda inside WSL, while the automation client may run inside WSL or on the Windows host; WSL forwards localhost port 9222.

Use a glibc-based Linux environment because the release binaries do not run directly on musl-based distributions such as Alpine. The documented binaries require glibc 2.35 on x86_64 and 2.38 on arm64; Android and Termux do not have a native build.

Related MCP servers

Add a badge to your READMEShow it’s listed — live SVG, Markdown or HTML.

AllMCPs Directory Badge

Full Badge Customizer

Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.

Badge Style:
Live Dynamic SVG PreviewBrowser AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/lightpanda-io-browser?style=directory)](https://allmcps.com/mcp/lightpanda-io-browser)
HTML Embed
<a href="https://allmcps.com/mcp/lightpanda-io-browser"><img src="https://allmcps.com/api/badge/lightpanda-io-browser?style=directory" alt="Browser on AllMCPs" /></a>