Search1API for LangChain

Official Search1API tools for LangChain agents. The package exposes live web search, news search, and page crawling through LangChain's standard tool interface while relying on the official Search1API Python SDK for authentication, retries, timeouts, and API errors.

The similarly named langchain-search1api package on PyPI is an independent third-party project. search1api-langchain is the official package maintained by Search1API.

Install

pip install search1api-langchain

Set an API key:

export SEARCH1API_API_KEY="your-api-key"

SEARCH1API_KEY is also accepted for compatibility with the Search1API CLI. Create a key in the Search1API dashboard.

Use all tools

from search1api_langchain import Search1APIToolkit

tools = Search1APIToolkit().get_tools()

search = tools[0]
result = search.invoke(
    {
        "query": "latest LangChain agent releases",
        "max_results": 5,
        "time_range": "month",
    }
)
print(result)

The toolkit returns:

• search1api_search for discovering ranked web sources;
• search1api_news for recent news articles;
• search1api_crawl for reading a known URL as clean content; and
• search1api_ask for automatic source routing and relevance ranking.

Pass the returned list directly to a LangChain agent:

from langchain.agents import create_agent
from search1api_langchain import Search1APIToolkit

agent = create_agent(
    model="openai:gpt-5.4",
    tools=Search1APIToolkit().get_tools(),
)

response = agent.invoke(
    {"messages": [{"role": "user", "content": "What changed in LangChain this month?"}]}
)

The agent example also requires the langchain package and the integration package for the selected model.

Use tools individually

from search1api_langchain import (
    Search1APICrawlTool,
    Search1APINewsTool,
    Search1APISearchTool,
)

search = Search1APISearchTool()
news = Search1APINewsTool()
crawl = Search1APICrawlTool()

search.invoke({"query": "agent observability", "search_service": "google"})
news.invoke({"query": "AI regulation", "time_range": "day"})
crawl.invoke({"url": "https://example.com/article"})

Every tool supports LangChain's asynchronous invocation:

result = await search.ainvoke({"query": "async Python agents"})

An explicit key and client settings can be supplied when environment-based configuration is not appropriate:

tool = Search1APISearchTool(
    api_key="your-api-key",
    timeout=45,
    max_retries=3,
)

API keys use Pydantic's secret type and are excluded from model serialization.

Credits

Plain Search and News requests cost one credit. Setting crawl_results asks Search1API to read the top result pages and costs one additional credit for each successful crawl. The tools therefore default crawl_results to 0.

A Crawl request costs one credit. See Credits and limits for current details.

Development

python -m venv .venv
. .venv/bin/activate
python -m pip install -e ".[dev]"
ruff format --check .
ruff check .
mypy src/search1api_langchain
pytest
python -m build

License

MIT

Current API capabilities

Search supports bingcn, yandex, and grokipedia; the removed sogou engine is no longer offered. Search and news support day, week, month, and year time ranges. Search also accepts page on engines with native pagination.

Agentic Search (ask) accepts only a natural-language query. The server chooses the engines and time window and returns query, intent, ranked results, and errors. Each Ask result includes its source and relevance. It returns search results, not a generated answer, and costs 5 credits per call. Use regular search when the engine and keywords are already known.

Search/news engine-specific attributes live inside each result's optional metadata: GitHub uses kind, stars, language, and num_comments; Hacker News news results use story_url, points, and num_comments. published_date remains optional at the result top level. Numeric zero counts are preserved; unavailable attributes are omitted. When displaying Grokipedia results, credit them as "Powered by xAI".

from search1api_langchain import Search1APIAskTool

result = Search1APIAskTool().invoke({"query": "recent papers on speculative decoding"})

The toolkit includes search1api_ask by default. ask_timeout defaults to 45 seconds independently of the other tools' timeout. Both synchronous and asynchronous invocation are supported. Requires search1api>=0.4.