LangChain | Firecrawl

Official package: langchain-firecrawl · sourceScrape, crawl, map, extract, and search from LangChain, as a document loader or as agent tools

Install

pip install langchain-firecrawl

Get a key from firecrawl.dev and set it in your environment (or pass api_key=...):

export FIRECRAWL_API_KEY="fc-your-api-key"

Document loader

FirecrawlLoader returns LangChain Documents. Choose a mode:

Mode	What it loads
`scrape`	One page
`crawl`	A whole site
`map`	Discovered URLs
`extract`	Structured data
`search`	Web search results

from langchain_firecrawl import FirecrawlLoader

loader = FirecrawlLoader(url="https://www.firecrawl.dev", mode="scrape")
docs = loader.load()

print(docs[0].page_content[:200])
print(docs[0].metadata)

Agent tools

Each capability is also a BaseTool you can bind to an agent:

from langchain_firecrawl import (
    FirecrawlScrape,
    FirecrawlCrawl,
    FirecrawlMap,
    FirecrawlExtract,
    FirecrawlSearch,
)

scrape = FirecrawlScrape()
result = scrape.invoke({"url": "https://www.firecrawl.dev"})
print(result["markdown"])

search = FirecrawlSearch()
print(search.invoke({"query": "best web scraping libraries", "limit": 5}))

LangChain JS

The FireCrawlLoader in @langchain/community still calls v1 SDK methods, so it fails with the current Firecrawl JS SDK. Until it’s updated, call the Firecrawl Node SDK directly and wrap the results in LangChain Documents:

npm install firecrawl @langchain/core

import Firecrawl from "firecrawl";
import { Document } from "@langchain/core/documents";

const firecrawl = new Firecrawl(); // reads FIRECRAWL_API_KEY

const crawl = await firecrawl.crawl("https://firecrawl.dev", { limit: 5 });

const docs = crawl.data.map(
  (page) => new Document({ pageContent: page.markdown ?? "", metadata: page.metadata ?? {} })
);

Resources