langchain-context

This package contains the LangChain integration for Context.dev, an API for turning the live web into reliable data for AI applications. It gives agents native tools for web search, scraping, crawling, structured extraction, document parsing, brand intelligence, website monitoring, screenshots, and asynchronous batch jobs.

Quick install

pip install -U langchain-context

Get an API key from context.dev and set it as the CONTEXT_API_KEY environment variable, or pass api_key=... to any tool or toolkit.

export CONTEXT_API_KEY="your-api-key"

Tools

Each Context.dev capability is available as a LangChain BaseTool:

from langchain_context import ContextScrape, ContextSearch

search = ContextSearch()
results = search.invoke(
    {
        "query": "latest official Stripe product announcements",
        "includeDomains": ["stripe.com"],
        "numResults": 10,
    }
)

scrape = ContextScrape()
page = scrape.invoke({"url": "https://stripe.com/newsroom"})

The most common tools have short aliases:

Tool	Purpose
ContextSearch	Search the live web
ContextScrape	Scrape one URL as clean Markdown
ContextCrawl	Crawl linked pages from a website
ContextSitemap	Discover or search a site's URLs
ContextExtract	Extract structured data using a JSON schema
ContextParse	Parse PDFs and other documents into Markdown

Dedicated classes are also available for HTML, images, screenshots, style guides, fonts, company classifications, brand profiles, monitors, and batch jobs.

Toolkit

Give an agent the complete safe-by-default Context.dev toolkit:

from langchain_context import ContextToolkit

tools = ContextToolkit().get_tools()

Or keep the tool list focused:

from langchain_context import (
    ContextBatchToolkit,
    ContextBrandToolkit,
    ContextMonitorToolkit,
    ContextWebToolkit,
)

web_tools = ContextWebToolkit().get_tools()
brand_tools = ContextBrandToolkit().get_tools()
monitor_tools = ContextMonitorToolkit().get_tools()
batch_tools = ContextBatchToolkit().get_tools()

Write operations and browser actions are opt-in. Enable them only for agents that should be allowed to create or modify Context.dev resources or interact with forms on third-party websites:

tools = ContextToolkit(
    include_write_tools=True,
    allow_browser_actions=True,
).get_tools()

All toolkits accept api_key, api_base, and timeout. api_base defaults to https://api.context.dev/v1 and can also be set with CONTEXT_API_BASE.

Parse documents

ContextParse accepts base64-encoded bytes so files remain compatible with LangChain's JSON tool-call interface:

import base64
from pathlib import Path

from langchain_context import ContextParse

file_bytes = Path("report.pdf").read_bytes()
result = ContextParse().invoke(
    {
        "fileBase64": base64.b64encode(file_bytes).decode("ascii"),
        "fileName": "report.pdf",
    }
)

Async

Every tool supports LangChain's asynchronous interface:

from langchain_context import ContextSearch

result = await ContextSearch().ainvoke(
    {"query": "latest Context.dev product announcements"}
)

Coverage

The package exposes all 33 API-backed tools in Context.dev's public MCP catalog. The MCP-only get-brand visual card is host UI; ContextBrandRetrieve returns the same underlying brand data as structured output for LangChain applications.

Documentation

• LangChain integration source
• Context.dev documentation
• Context.dev API reference
• Context.dev homepage