What result fields are returned for LLM context?

Each organic result includes: position (rank position), title (page title), link (full URL), snippet (text excerpt). These fields map directly to LLM context without any parsing or transformation.

AI Agent & Copilot Development

Web Search API for AI Agents

Stop your agents from hallucinating. Give them live web access through one REST call: structured JSON results ready for any LLM context window.

Get Started Free Try the Playground View Pricing

50M+

API requests served

99.9%

Uptime SLA

100+

Countries supported

<5s

Median response time

Trusted by companies worldwide

Capabilities

What you can build with a search-enabled agent

The Web Search API gives your agents a live window into the world. Here is what that unlocks.

Tool & Function Calling

Register SERPHouse as a tool in OpenAI function calling, Anthropic tool use, or any agent framework. One function definition gives your agent instant web access.

Hallucination Prevention

Force agents to search before they answer. Ground every response in current, verified web data instead of training memory from months ago.

Multi-Step Research

Chain searches in an agent loop. Each result informs the next query. Build agents that conduct deep, systematic research across multiple sources.

Always Current Knowledge

No knowledge cutoff. No stale context. Your agent searches the live web and knows what happened today.

Parallel Search Execution

Run multiple search queries concurrently in async agent pipelines. Surface broad context across topics, regions, and sources in a single agent turn.

Structured LLM-Ready Output

Every result returns position, title, url and snippet. Drops directly into your prompt, with no HTML parsing or preprocessing required.

How It Works

From zero to a grounded AI agent in 4 steps

1

Get your API key

Sign up free at SERPHouse. No credit card required. Copy your API key from the dashboard in under 2 minutes.

2

Register as an agent tool

Define a search function that calls the SERPHouse REST API. Pass it to your agent framework as a tool or function that works with OpenAI, Anthropic, LangChain, LlamaIndex, LangGraph, CrewAI, and AutoGen.

3

Agent triggers a search

When the agent needs current information, it calls the search tool with a query. The API responds in under 1 second with structured JSON including organic results, each with position, title, url and snippet.

4

Ground the LLM response

Inject the search results into the LLM system prompt as context. The model reasons over real data, not hallucinated training memory.

Why grounding matters

Without grounding

Agent answers from training data with a knowledge cutoff
Hallucination risk on recent events or current facts
No citations, user cannot verify the source

With SERPHouse grounding

Agent retrieves live facts before responding
Accurate, current, and citable answers
URLs returned with each result, resulting in fewer hallucinations, better trust

Result: Fewer hallucinations, better user trust, and citable search results with real URLs your users can verify.

Code Examples

Register search as an agent tool in minutes

One REST endpoint. Standard bearer auth. The structured JSON response drops directly into your agent context with no transformation required. Works with OpenAI, Anthropic Claude, LangChain, LlamaIndex, LangGraph, CrewAI, and AutoGen.

GET /serp/live · REST endpoint, no SDK needed

Authorization: Bearer · Standard bearer token auth

results.organic[] · position,title, link, snippet per result

Get API Key Free Try Playground

import requests
from openai import OpenAI

def search_web(query, num_result=5):
    """Search the live web; register this as an AI agent tool."""
    response = requests.get(
        "https://api.serphouse.com/serp/live",
        headers={"Authorization": "Bearer YOUR_API_KEY"},
        params={"q": query,
                "loc": "United+States", "num_result": 5}
    )
    return response.json()["results"]["organic"]

# Register with OpenAI function calling
client = OpenAI()
tools = [{
    "type": "function",
    "function": {
        "name": "search_web",
        "description": "Search the live web for current information.",
        "parameters": {
            "type": "object",
            "properties": {
                "query": {"type": "string",
                         "description": "The search query"},
                "num_result": {"type": "integer", "default": 5}
            },
            "required": ["query"]
        }
    }
}]

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user",
               "content": "Latest AI agent frameworks in 2025?"}],
    tools=tools,
    tool_choice="auto"
)

const API_KEY = 'YOUR_API_KEY';

async function searchWeb(query, numResults = 5) {
  const params = new URLSearchParams({
    q: query,
    loc: 'United+States'
  });
  const res = await fetch(
    'https://api.serphouse.com/serp/live?' + params.toString(),
    { headers: { 'Authorization': 'Bearer ' + API_KEY } }
  );
  const data = await res.json();
  // [{ position, title, link, snippet }]
  return data.results.organic;
}

// Register as a tool in any agent framework
const agentTools = [{
  name: 'search_web',
  description: 'Search the live web for current information',
  execute: searchWeb
}];

curl -X GET \
  "https://api.serphouse.com/serp/live?q=latest+AI+agent+frameworks+2025&loc=United+States&num_result=5" \
  -H "Authorization: Bearer YOUR_API_KEY"

{
  "results": {
    "search_information": [
      {
        "total_results": "About 98,400,000 results",
        "time_taken_displayed": "0.19"
      }
    ],
    "organic": [
      {
        "position":     1,
        "title":   "Top AI Agent Frameworks in 2025: LangChain, AutoGen, CrewAI",
        "link":     "https://example.com/ai-agent-frameworks-2025",
        "snippet": "A comprehensive guide to the leading AI agent frameworks..."
      }
    ]
  }
}

Try it live in the API Playground →

Why SERPHouse

One search API. Every agent framework.

Most web search APIs are designed for traditional applications. SERPHouse is built with AI agent pipelines in mind: structured JSON output, sub-second latency, and 100+ country targeting means your agents can research any topic, in any region, without specialized infrastructure.

Works with OpenAI, Anthropic, Google Gemini function calling

Compatible with LangChain, LlamaIndex, LangGraph, CrewAI, AutoGen

REST API: no SDK required, any HTTP client works

Async-ready for concurrent multi-query agent turns

Production SLA: 99.9% uptime with automatic retry handling

OpenAI

Register as a function tool in gpt-4o, o3, or any OpenAI model. The structured JSON response maps directly to tool output for the model to reason over.

Anthropic Claude

Use as a tool in Claude 3/4 tool use API. Define the search tool schema once and handle the tool_use content block in your response loop.

LangChain / LlamaIndex

Subclass BaseTool in LangChain or create a custom LlamaIndex reader. See the full integration guide at LangChain Integration.

FAQ

Frequently asked questions

Everything developers ask before adding web search to their AI agents.

Register a function that calls the SERPHouse REST API: GET https://api.serphouse.com/serp/live with your query and API key. Pass this function as a tool to your agent framework (OpenAI function calling, LangChain Tool, Anthropic tool use). When the agent needs current information, it calls the function and receives structured JSON results it can include in its response.

Yes. Define a Python or JavaScript function that makes a GET request to the SERPHouse API, then register it as a tool in your OpenAI client call using the tools parameter. When the model decides to search, it triggers your function and receives results as structured JSON (including title, link, snippet, and position) for each result.

Yes. The SERPHouse Web Search API is a standard REST endpoint that works with any tool use implementation. Define the tool schema in your Anthropic client call, implement the function to call the SERPHouse API, and handle the tool_use content block in the response.

Ground agent responses in live web data by requiring the agent to search before answering factual questions. When an agent retrieves current web results via SERPHouse and uses them as context, it generates answers based on real, citable information rather than training memory. This is especially important for time-sensitive queries, recent events, and rapidly changing facts.

The SERPHouse API returns results in under 5 seconds on average (median 48ms for simple queries). In an agent pipeline, this adds one network round trip per search. For async agents running parallel tool calls, multiple searches add no additional wall-clock time. The latency is negligible compared to the LLM inference time.

Each organic result includes: position (rank position), title (page title), url (full URL), snippet (text excerpt). These fields map directly to LLM context without any parsing or transformation.

SERPHouse rate limits depend on your plan. Free tier supports limited requests per month for testing. Paid plans support higher concurrency for production agent pipelines. For high-volume agent applications making dozens of searches per agent turn, enterprise plans provide dedicated rate limits and SLA-backed uptime. See pricing for details.

Yes. Pass the date_range parameter to filter by time range (such as date_range=h for the past hour, date_range=d for past 24 hours, date_range=w for past week. This is essential for agents that need breaking news, current prices, or recent events. Combine with the loc parameter for country-specific recency filtering).

Related Use Cases

More ways to build with the SERPHouse APIs

RAG Pipelines

Use live web results as retrieval context in your RAG system. No stale vector store, always current.

Learn more

LangChain & LlamaIndex Integration

Drop-in BaseTool and LlamaIndex reader for framework-based agents. Full code example included.

Learn more

Real-Time News Monitoring

Keep agents informed with live news search results. Filter by topic, recency, and region.

Learn more

Start building smarter AI agents today

Free tier available. No credit card required. Your agent makes its first live web search in under 5 minutes.

Get Started Free Try the Playground Contact for Enterprise