Build Production-Ready AI Solutions with You.com's Express API and Custom Agents

Build Production-Ready AI Solutions with You.com's Express API and Custom Agents
10:49

Large Language Models (LLMs) are incredibly powerful, but using them in a production environment highlights a critical challenge: they are prone to errors and lack real-time knowledge of the world. For AI applications and enterprise solutions where accuracy, reliability, and trust are non-negotiable, companies and developers need more than just a base model.

This is where you.com’s Express API and Custom Agents (CAs) come in. We provide composable AI infrastructure that enhances leading LLMs with a state-of-the-art web search layer, giving you the power to build fast, factual, and production-ready AI solutions. The Express API is the fastest way to get answers enhanced by our web search capabilities, and the CA API gives you the flexibility to choose your base model from over 20+ models from various providers such as OpenAI, Anthropic, Gemini, and more. You.com updates model availability as soon as providers release them, giving you the flexibility to always use the most capable models for your workflows.

Our core philosophy is that for an AI application to be production-ready, its answers must be accurate, trustworthy, and grounded in the latest information. A base LLM, no matter how powerful, is a closed system that hallucinates and lacks real-time knowledge. That's why we prioritize factual, up-to-date answers by ensuring all foundational models on our platform are enhanced with web search. This isn't an optional add-on; it's a fundamental part of our architecture. For example, our platform supports gpt-5-mini with web search, which OpenAI does not currently offer for that model.

Benchmark Criteria & Evaluation 

We evaluate our systems on SimpleQA, a benchmark that measures accuracy and F1 scores for real-world question-answering tasks. The results show that our CAs deliver state-of-the-art accuracy with industry-leading speed.

Latency is a critical factor for user experience. We measure it in percentiles:

  • p50 (Median): The typical experience for most users.
  • p95: The experience for your power users or more complex queries.

SimpleQA Dataset & Findings

Since OpenAI does not offer web search with gpt-5-mini we decided to evaluate using gpt-4.1-mini as base model for a fair comparison:

You.com achieves 2.3x better accuracy at 1/5th the cost through our proprietary search infrastructure, optimized web retrieval algorithms and prompt engineering. The 2-3 second latency trade-off transforms gpt-4.1-mini from 26% to 58% accuracy, proving You.com's differentiation extends beyond being model agnostic to our underlying search and AI infrastructure. 

Metric You.com express CA w/ gpt-4.1-mini gpt-4.1-mini w/ web search Improvement
Simple QA Accuracy 0.583 0.258 +126%
Simple QA F1 0.596 0.259 +130%
Cost per Query (outside of the input/output tokens) $0.005 $0.025 5x less expensive
Latency p50 3.88s 1.58s Trade-off

 

For a full comparison refer to the following table: 

 

latency p50

latency p95

simple QA accuracy

simple QA f1

cost/query (outside of input/output tokens)

you.com express CA w/ gpt-4.1-mini

3.88

5.48

0.583

0.596

$0.005

you.com express CA w/ gpt-4.1

3.31

4.84

0.587

0.601

$0.005

you.com express CA w/ gpt-5

3.36

5.11

0.585

0.598

$0.005

you.com express CA w/ gpt-5-mini

3.45

5.74

0.587

0.602

$0.005

you.com express

2.57

4.17

0.584

0.591

$0.005

gpt-4.1-mini w/ web search

1.58

4.02

0.258

0.259

$0.025

The benchmark results prove that You.com prioritizes delivering reliable, high-quality answers. The small trade-off in speed in exchange for significantly better accuracy makes You.com’s models a smart choice for users who value precision at a fifth of the cost. 

 

What is Possible Today

The Express API and CA API are designed to provide developers with an easy-to-integrate solution in their platform that will allow them to build solutions that require fast and factual answers for a variety of use cases. Here are some examples of how our Express API can be used:

  1. Enhanced Customer Support & Virtual Assistants

Build AI-powered chatbots and virtual assistants that deliver fast, accurate, and up-to-date answers by combining LLM reasoning with live web search results. This dramatically improves customer experience across industries like retail, telecom, and SaaS by reducing response times and increasing factual accuracy.

  1. Dynamic Knowledge Management & FAQ Systems

Create intelligent knowledge bases and FAQ tools that continuously update with fresh information from the web, complementing static internal data. This is valuable for organizations in education, tech support, HR, and more, enabling users to get relevant answers without manual content updates.

  1. Research & Content Generation with Fact-Checking

Assist researchers, journalists, and content creators by providing AI-generated summaries and insights augmented with real-time factual data from the web. This use case helps ensure content is both creative and grounded in the latest information, improving trustworthiness and relevance.

Create Your Own Custom Agents

You can create custom agents on our platform with any of our supported foundational models. To create one, simply navigate to you.com and select the “+” button in the agent sidebar. There, you can give your agent a name, description, prompt, and base model. All of the foundational models on the custom agent model selector are compatible with the CA API. Here are some examples of custom agents you can create on our platform, and how you can use the CA API to build products around these agents:

 

Name of Agent

Description of Agent

Example Use Case

Citation Finder

Finds credible sources supporting or refuting a user’s claim, returning concise citations.

Embed the agent’s API in a collaborative document editor. As users write, the editor sends highlighted text to the API, which returns the text with formatted citations and links, allowing users to insert references with one click.

Quick Comparison Agent

Instantly compares two products, services, or concepts using web summaries and reviews.

Add the agent’s API to an e-commerce product page. When a shopper clicks “Compare,” the site calls the API with two product IDs, and displays a side-by-side comparison table with summarized pros, cons, and recent reviews.

Event Verifier Agent

Verifies if a specific event has occurred, providing a yes/no answer with a source link.

Integrate the agent’s API into a financial trading platform. When a user sets up a news-based trading trigger, the platform queries the API to confirm if the event (e.g., “Company X announced earnings”) has actually happened before executing trades.

Local Alert Agent

Provides real-time, location-based alerts for weather, emergencies, or transit.

Integrate the agent’s API into a travel planning app. When a user books a trip, the app will make a call to the API for that location, pushing instant notifications about severe weather, transit delays, or emergencies to the user’s device.

Example Public Custom Agents

Here are some examples of public express agents and custom agents you can try querying today:

CA Name Base Model Description UUID for Payload
Counterpoint Agent Gemini 2.5 Pro Your AI thought partner that challenges assumptions, identifies blind spots, and tests reasoning in business documents to strengthen analysis. Start with a greeting ('hi', 'hello'). 998a97e2-a8cf-457c-b5c3-8fd109fb009e
Presentation Slide Generator Gemini 2.5 Pro Create tailored presentations, uncover target audience, develop outlines, draft content, apply voice and tone, get feedback, and generate presentation content. Start with a greeting ('hi', 'hello'). 18969237-b534-4dbe-9a7c-c2c9370a8685
SEO Content Crafter Claude 3.5 Sonnet What keyword are you trying to optimize? 8d28d616-95b1-4f48-8e43-5cde8c540f8f

Limitations

  • As of now, the CA API returns a 403 error if your agent requests a tool that it or the tenant is not authorized to use. Make sure to synchronize the agent’s allowlist with the tools it requests. Providing an empty array disables tools for the current run. 
  • Custom Agents configured with Advanced Reasoning or Research modes are not supported via the API.
  • Files attached to a Custom Agent are not accessible when the agent is invoked via the API. 

Get Started in Minutes

Ready to build more reliable AI applications and simplify your AI stack? You can make your first call to any major LLM through our unified API in just a few steps.

  1. Create a Custom Agent: Go to you.com, create a new CA, and configure its prompt and sources.
  2. Get the Agent UUID: Copy the UUID from the agent's URL. The URL will look like ...chatMode=user_mode_[UUID HERE].
  3. Make Your API Call: Use the UUID in your request payload for the agent parameter. To use our default Express agent, simply use "express" as the value.

Appendix


Here is an example request: 

curl -X POST https://api.you.com/v1/agent/runs \

  -H "Content-Type: application/json" \

  -H "Authorization: Bearer YOUR_API_KEY" \

  -d '{

  "agent": "cd0b708d-9135-485e-be46-51153606c20e", 

  "input": [

    {

      "content": "whats the latest news on ai?" 

    }

  ],

  "tools": [

    {"type": "web_search"} 

  ],

  "stream": false, 

  "store": false 

}'

 

Start building more accurate, reliable, and future-proof AI applications today.

For more details on API request formatting and other parameters, please refer to our Developer Documentation.