unbrowse-langchain 2.1.4

LangChain tools for the Unbrowse CLI — reverse-engineer any website into reusable API skills.

unbrowse-langchain

LangChain tools for Unbrowse — drop-in replacement for PlayWrightBrowserToolkit and WebBrowserTool that returns structured JSON instead of raw HTML.

Bootstrap Unbrowse first

npx unbrowse setup
unbrowse health

Installation

npx unbrowse setup
pip install unbrowse-langchain

Quick Start

Replace PlayWrightBrowserToolkit

Before (unstructured HTML, slow, needs Chromium):

from langchain_community.agent_toolkits import PlayWrightBrowserToolkit
from langchain_community.tools.playwright.utils import create_sync_playwright_browser

browser = create_sync_playwright_browser()
tools = PlayWrightBrowserToolkit.from_browser(browser).get_tools()

After (structured JSON, cached, no browser needed):

from unbrowse_langchain import create_unbrowse_toolkit

tools = create_unbrowse_toolkit()

With a ReAct Agent

from langchain_openai import ChatOpenAI
from langgraph.prebuilt import create_react_agent
from unbrowse_langchain import create_unbrowse_toolkit

llm = ChatOpenAI(model="gpt-4o")
tools = create_unbrowse_toolkit()
agent = create_react_agent(llm, tools)

result = agent.invoke({
    "messages": [("user", "Get the top stories from Hacker News")]
})

The agent will call unbrowse_resolve with the URL and intent, getting back structured data like:

[
  {"title": "OpenCode", "link": "https://opencode.ai/", "score": "580 points", "rank": "1"},
  {"title": "Molly Guard", "link": "https://...", "score": "38 points", "rank": "2"}
]

With Anthropic

from langchain_anthropic import ChatAnthropic
from langgraph.prebuilt import create_react_agent
from unbrowse_langchain import create_unbrowse_toolkit

llm = ChatAnthropic(model="claude-sonnet-4-20250514")
tools = create_unbrowse_toolkit()
agent = create_react_agent(llm, tools)

result = agent.invoke({
    "messages": [("user", "Get the trending repos on GitHub")]
})

Individual Tools

from unbrowse_langchain import UnbrowseResolveTool, UnbrowseSearchTool

# Resolve a website into structured data
resolve = UnbrowseResolveTool()
result = resolve.invoke({
    "intent": "get top stories",
    "url": "https://news.ycombinator.com",
})
# Returns: [{"title": "...", "link": "...", "score": "..."}, ...]

# Search the skill marketplace
search = UnbrowseSearchTool()
result = search.invoke({"intent": "weather forecast api"})

Custom Configuration

tools = create_unbrowse_toolkit(
    bin_path="/opt/bin/unbrowse",   # custom binary path
    timeout=60,                     # timeout in seconds (default: 120)
)

How It Replaces the Browser

	PlayWrightBrowserToolkit	unbrowse-langchain
Output	Raw HTML/text, agent must parse	Structured JSON with extracted fields
Speed (first)	5-30s per page	5-15s (discovers + caches API)
Speed (cached)	Same every time	300ms-1s
Dependencies	Chromium, Playwright	unbrowse CLI only
Auth support	Manual cookie injection	unbrowse login --url captures session
Reliability	Breaks on DOM changes	Uses discovered API endpoints

Available Tools

Tool	Description	Required args
unbrowse_resolve	Reverse-engineer a page into structured data	intent, url
unbrowse_search	Search the skill marketplace	intent
unbrowse_execute	Execute a cached skill endpoint	skill_id, endpoint_id
unbrowse_login	Authenticate with a site	url
unbrowse_skills	List cached skills	--
unbrowse_skill	Inspect a skill	skill_id
unbrowse_health	Health check	--

Development

cd integrations/langchain
pip install -e ".[dev]"
pytest tests/   # 14 tests