Toolbox/Agent capabilities

Browser agents

browser automation and web agents

8 repositories, most starred first

/

State of the category

firecrawl/firecrawl firecrawl/firecrawl is the clear category leader — it combines large-scale crawling with an agent-ready web context API that converts JS-heavy pages into LLM-ready Markdown/JSON and supports actions. The single most useful thing to know: pick Firecrawl for scalable, agent-oriented scraping and action execution; pick something else only if you need different runtime/privacy/performance tradeoffs.

Local privacy and BYO

Many teams require browsers and agents that run fully under their control and can use bring‑your‑own models for compliance, latency, or cost reasons, which hosted, centralized scrapers don't satisfy.

→ browser-use, OpenCLI, moli

In-page GUI

Product teams often want an agent that embeds directly into a page with a single script so end‑users get interactive automation without extra infra or headless browsers.

→ page-agent

Reproducible automation

Engineering teams need deterministic, scriptable outputs (reusable CLI/Playwright commands or learned navigation) for CI, debugging, and predictable cost control instead of ad‑hoc agent steps.

→ OpenCLI, Webwright, webcmd

Lightweight rendering

Resource‑constrained deployments and CI environments benefit from a DOM‑first, non‑Chromium engine that reduces memory and CPU compared with full Chromium stacks.

→ moli

Real‑browser bridging

Automating authenticated workflows without exposing credentials requires tooling that drives a real Chrome tab, preserves login state, and can mask sensitive fields.

→ dsh-browser, OpenCLI

Find your fit

Which one matches your setup?

Answer any of the questions — the shortlist updates as you go. Recommendations come from the capability passports below, nothing else.

Where should reviews happen?

Can code leave your infrastructure?

What matters most?

Model access?

Pick at least one answer to get a shortlist.

Side by side

Comparison matrix

Axes are extracted from each project's docs by our review pipeline; the maturity score is computed from stars, growth and commit activity — not an opinion. Click a column to sort.

	Runs in	Models	Context					Cost to run
firecrawl/firecrawl ↗Leader ⭐ 176.3k +2902/7d	CLICoding-agent pluginWeb app	Fixed provider	Whole-site crawling	Structured output	●●●●●	Hosted API key	●●●●●	Hosted service — API key required
browser-use/browser-use ↗ ⭐ 112.2k +704/7d	CLICoding-agent plugin	BYOKOpenAIAnthropicGeminiOllama / localFixed provider	Diff only	None mentioned	●●●●●	Fully local capable	●●●●●	Your LLM API key (or run free local models).
alibaba/page-agent ↗ ⭐ 29.0k +91/7d	Web app	BYOKOllama / local	Single-page only	No noise controls	●●●●●	Self-hostable	●●●●●	Demo testing LLM available; otherwise use your API key
jackwener/OpenCLI ↗ ⭐ 29.0k +286/7d	CLICoding-agent plugin	AnthropicGeminiFixed provider	Site & related files	Basic diagnostics	●●●●●	Runs locally	●●●●●	Free, local runtime
microsoft/Webwright ↗ ⭐ 6.0k +7/7d	CLICoding-agent pluginWeb app	OpenAIAnthropicFixed provider	Workspace-wide	Screenshot verification	●●●●●	Cloud APIs	●●●●●	Your API key for chosen backend (OpenAI / Anthropic / OpenRouter).
lexmount/moli ↗ ⭐ 1.7k +357/7d	CLICI	Fixed provider	No repo insight	Opt-in controls	●●●●●	Runs locally	●●●●●	Free, local binary
agentrhq/webcmd ↗ ⭐ 1.6k	CLICoding-agent pluginWeb app	OpenAIAnthropicGeminiFixed provider	Diff only	Minimal controls	●●●●●	Self-hostable	●●●●●	Free to run locally; optional hosted cloud
Lum1104/dsh-browser ↗ ⭐ 570 +75/7d	CLIWeb appCoding-agent plugin	BYOK	Current page only	User approvals & masking	●●●●●	Local harness, external model	●●●●●	Free to install; model usage may incur provider costs

At a glance

Capability profiles

Six axes, 0–5 each. The shape tells you the strategy: a wide hexagon is a generalist, a spike is a specialist.

firecrawl/firecrawl⭐ 176.3k

Provides an agent-ready web context API that scrapes JS-heavy pages into LLM-ready Markdown/JSON and supports actions (click/scroll/write) plus large-scale crawling — a combination many scrapers don’t offer.

browser-use/browser-use⭐ 112.2k

An open-source browser agent that runs locally and is extensible with custom tools and provider‑prefixed models, enabling reproducible, privacy-friendly web automation.

alibaba/page-agent⭐ 29.0k

An in-page GUI agent that runs entirely inside any web page with a single script, no extension or headless browser required.

jackwener/OpenCLI⭐ 29.0k

Converts any website into a deterministic CLI and lets AI agents drive your logged-in Chrome via a local Browser Bridge and reusable adapter skills.

microsoft/Webwright⭐ 6.0k

Treats the browser as a disposable environment and the local workspace (code, logs, screenshots) as the persistent state, producing reusable Playwright scripts rather than single-step agent actions.

lexmount/moli⭐ 1.7k

A standalone, Rust-built, DOM-first headless browser that performs on-demand layout and rendering to minimize resource use for AI agent workloads compared with Chromium-based engines.

agentrhq/webcmd⭐ 1.6k

Learns site navigation and compiles it into reusable deterministic commands so agents stop rediscovering the web and greatly reduce token spend.

Lum1104/dsh-browser⭐ 570

Provides a text-only bridge to your real Chrome tab, preserving login state and masking sensitive fields so the model can operate without screenshots or exposing secrets.