SerpApi Search Tools

Give Python AI agents live web, news, maps, image, shopping, video, hotel, and flight search with easy-to-use, customizable tools.

Read the full documentation for guides, SDK examples, recipes, and the API reference.

The package creates native SerpApi tools for popular Python agent SDKs:

from serpapi_search_tools import maps_search, news_search, web_search

tools = [
    web_search(),
    news_search(),
    maps_search(),
]

When one supported agent SDK is installed, the package detects it and creates tools ready for that SDK.

Install

If your agent SDK is already installed, add only the base package:

pip install serpapi-search-tools

If you want this package to install a compatible agent SDK too, choose its extra. For example:

pip install "serpapi-search-tools[openai-agents]"

Extras are available for all supported SDKs listed below.

Set a SerpApi key:

export SERPAPI_API_KEY="your-key"

SERPAPI_KEY is also supported. A directly supplied api_key= takes precedence over environment variables.

Supported agent SDKs

Quickstart: automatic SDK detection

This quickstart uses OpenAI Agents SDK to demonstrate automatic detection. It assumes the SDK is already installed in your environment (install it with pip install openai-agents if needed). Then add the base package:

pip install serpapi-search-tools

With one supported SDK installed, create the tool without any configuration. The package detects OpenAI Agents SDK and returns its native FunctionTool. This example also expects the OPENAI_API_KEY used by your agent.

from agents import Agent, Runner

from serpapi_search_tools import web_search

agent = Agent(
    name="research-agent",
    instructions="Use web search when the answer needs current information.",
    tools=[web_search()],
)

result = Runner.run_sync(
    agent,
    "Find three recent Python packaging changes and explain why they matter.",
)
print(result.final_output)

Quickstart: LangChain

Install the LangChain extra and the model backend used by this example:

pip install "serpapi-search-tools[langchain]" langchain-openai

The langchain extra installs a compatible LangChain version. langchain-openai provides this example's model integration; replace it with the backend your LangChain application uses. With langchain-openai, set OPENAI_API_KEY before running the agent.

from langchain.agents import create_agent
from langchain_openai import ChatOpenAI

from serpapi_search_tools import maps_search, news_search, web_search

agent = create_agent(
    model=ChatOpenAI(model="gpt-5.4-mini", temperature=0),
    tools=[
        web_search(),
        news_search(),
        maps_search(),
    ],
)

result = agent.invoke(
    {
        "messages": [
            {
                "role": "user",
                "content": (
                    "Research coffee culture in Austin using current reporting, "
                    "local places, and general web sources."
                ),
            }
        ]
    }
)
print(result["messages"][-1].content)

The constructors use automatic SDK detection, just as in the first quickstart. For multi-SDK environments and explicit selection, see Agent SDKs.

For a step-by-step explanation, keys, customization, and troubleshooting, read the detailed quickstart.

Browse the runnable examples for focused integrations or the agent cookbook for complete, task-oriented agents built with every supported SDK.

Choose the right tool

General web

from serpapi_search_tools import WebSearchEngine, web_search

tool = web_search(
    allowed_engines=[WebSearchEngine.GOOGLE_LIGHT, WebSearchEngine.BING],
    default_engine=WebSearchEngine.GOOGLE_LIGHT,
)

SerpApi supports multiple general web search engines, including Google Light, Google, Bing, Yahoo, and DuckDuckGo. Google Light is the default because it provides fast, general-purpose web results. Use allowed_engines to choose which engines are available and default_engine to select the initial one.

News, maps, images, and videos

from serpapi_search_tools import images_search, maps_search, news_search, videos_search

tools = [
    news_search(),
    maps_search(),
    images_search(),
    videos_search(),
]

news_search supports keyword searches in Google News. maps_search searches Google Maps and accepts optional location, zoom (3 through 30), and nearby fields. Use nearby=True for “near me” intent with a separate location; leave it false when the query already names a city or area. Place details, reviews, and directions use different SerpApi APIs and are not part of this search tool.

images_search returns images and their source pages. videos_search searches YouTube videos, Shorts, channels, playlists, movies, and categories.

Shopping

from serpapi_search_tools import ShoppingSearchEngine, shopping_search

tool = shopping_search(
    allowed_engines=[
        ShoppingSearchEngine.GOOGLE_SHOPPING,
        ShoppingSearchEngine.AMAZON,
        ShoppingSearchEngine.WALMART,
        ShoppingSearchEngine.EBAY,
    ],
)

Use the same query input for every marketplace. The package translates it to the selected engine's request format.

Travel

from serpapi_search_tools import flights_search, hotels_search, travel_explore_search

travel_tools = [
    hotels_search(),
    flights_search(),
    travel_explore_search(),
]

These constructors create hotel, flight, and destination-discovery tools for the detected agent SDK. Read the Hotels, Flights, and Travel Explore guides for their date, occupancy, location ID, and trip rules.

Configure tools

Every constructor accepts:

web_search and shopping_search also accept allowed_engines and default_engine. The tool offers only the engine values you configure.

Use default_params for documented SerpApi settings that should stay under your application's control, such as locale, currency, safe search, or pagination. Use result_limit to control how many results the tool returns. The agent continues to supply only the inputs described by its search tool.

Tools return Markdown by default. Markdown keeps links and tables readable without JSON syntax overhead. Application code that needs structured fields can opt into JSON:

import json

from serpapi_search_tools import SearchResultFormat, web_search

search = web_search(
    provider="function",
    response_format=SearchResultFormat.JSON,
)
result = json.loads(search(query="Python packaging"))

tool = news_search(
    default_params={"hl": "en", "gl": "us"},
)

Typed tool inputs override matching values in default_params, and the constructor always controls engine. The package rejects known incompatible combinations, such as a Google News query with a topic token, an Amazon keyword search with node, or flight airline include and exclude filters. A multi-engine tool sends the same defaults to every allowed engine, so use parameters shared by those engines or create separate tool instances. Reserved keys (api_key, async, engine, and output) are rejected in default_params; use the constructor options above instead.

Read Manage LLM context for compact and full response behavior, default result limits, and unlimited responses.

Handle search failures

The search runtime raises SerpApiSearchError for SerpApi and transport failures. Invalid tool inputs raise ValueError. Agent SDKs surface or handle tool errors differently, so use your SDK's normal tool-error mechanism. See Debugging for detailed examples.

Engine API references

For AI coding agents that need broader SerpApi API context, use SerpApi's agent-friendly documentation index (llms.txt). It links directly to Markdown API references, including APIs beyond those wrapped by this package.

More guides

• Full documentation
• SDK examples
• Agent cookbook