In this guide, we’ll pin Exa and Bright Data head to head. We’ll cover main offerings, pricing packages as well as a basic LangChain integration that teams can use.
By the time you’ve finished reading, you’ll be able to answer the following questions.
- What is Exa?
- What is Bright Data?
- What do these providers offer?
- What are their differences and similarities?
What exactly do Exa and Bright Data offer?
To be productive, AI agents need search and web access APIs. Whether you’re building a news agent, a trading agent or travel agent, your LLM needs to perform a search. After it performs a search, it needs to verify results.
- Search: Agents need to be able to search the web and find relevant results.
- Unlocking: Unlocking APIs give AI agents the tools to access full web pages to verify search results and gain greater context.
Exa
Exa is mainly thought of as a search engine. However, they offer a variety of tools that your AI agents can use. With Exa, teams can build agents and RAG pipelines using these tools.
- Search: Exa’s flagship product. Get AI-native search results with highlights and highlight scoring so agents can better understand the results.
- Contents: Read full web page text to verify search results and retrieve detailed information.
- Answer: Ask a question and get an AI-generated answer.
- Websets: Using Exa’s agentic search platform, generate web datasets with a single prompt.
- Monitors: Monitor the web in real time and receive updates via webhooks.
Bright Data
Bright Data home page
Bright Data offers a comprehensive suite of data tools. Their offerings range from proxies and web unblocking to web scraping APIs and datasets. Teams choose Bright Data when they’re looking for an enterprise suite of tooling built to handle a variety of data needs.
- Proxies: Route your web traffic through your choice of Internet Service Provider (ISP), residential and datacenter proxies.
- Unlocker API: Automated proxy rotation and CAPTCHA handling. Build software with stable web access.
- Discover API: Feed a ranked list of public URLs into your data pipeline. Discover trends as they’re taking shape.
- SERP API: Get structured search results from Google, Bing, DuckDuckGo and Yandex.
- Browser API: Run cloud browsers for stable and efficient browser automation. These are fully hosted with proxy rotation and CAPTCHA handling.
- Scraper APIs: Use prebuilt web scrapers to retrieve on demand data.
- Scraper Studio: An Integrated Development Environment (IDE) for building web scrapers supporting both traditional and no-code options.
- Datasets: Use curated datasets for analysis of macro trends and AI training.
- Data Firehose: Receive real-time data using custom built pipelines.
Pricing and usage limits
Exa
Exa pricing page
Exa offers a variety of pricing packages based on project needs. Most of these offerings are priced significantly higher than their counterparts from other providers. However, the straightforward pricing can be advantageous for teams who need predictable costs.
- Search: $7/1,000 requests.
- Deep Search: $12-$15/1,000 requests based on complexity.
- Contents: $1/1,000 requests.
- Monitors: $15/1,000 requests.
- Answers: $5/1,000 requests.
Bright Data
Bright Data pricing
Bright Data offers volume based pricing for each of their products. As your usage gets higher, your price per request or per GB gets lower.
- Proxies: Residential and ISP start at $8/GB (pay-as-you-go) and scales to $5/GB at high usage. ISP proxies also offer per-IP based pricing starting at $1.80/month per IP to $1.30/month per IP for shared proxies. Datacenter proxies range from $1.40/month per IP to $0.90/month per IP for shared options. Bandwidth-based datacenter proxies begin at $0.60/GB and go as low as $0.42/GB at scale.
- Unlocker API: Starts at $1.50/1,000 requests and scales to $1/1,000 requests (1,000,000 requests).
- Discover API: Currently in free beta.
- SERP API: Starts at $1.50/1,000 requests (pay-as-you-go) and scales to $1/1,000 requests (1,000,000 requests).
- Browser API: Starts at $8/GB (pay-as-you-go) and scales to $5/GB (399GB).
- Scraper API: Starts at $1.50/1,000 records. The highest tier usage costs $499/month (384,000 records). Custom enterprise pricing is available as well.
- Scraper Studio: Begins at $1.50/1,000 page loads. Scales to $1/1,000 page loads (2,000,000 pages loads total).
- Datasets: Pricing starts at $2.50/1,000 records (100,000 total) for one-time access. At high usage, teams pay $1/1,000 (minimum 5,000,000 records).
- Data Firehose: Fresh pages cost $0.20/1,000 HTML pages. Historical pages cost $1/1,000 HTML pages.
Usage examples
Now, let’s try building an AI agent using LangChain and each provider. We’ll be using the LangChain Exa package as well as LangChain Bright Data. Basic familiarity Python is required to use these tools with LangChain. Teams should note that the LangChain integrations do not provide full access to either company’s full offerings. You can learn more about wrapping LangChain tools here.
Getting started
Create a new virtual environment.
python -m venv .venvActivate the environment. The command below will activate for a Windows environment.
..venvScriptsActivate.ps1For Linux or macOS, use this activation command instead.
source .venv/bin/activateNow, we’ll install our dependencies. We’ll use langchain, langchain-openai, langchain-exa, langchain-brightdata and langgraph.
pip install langchain-openai langchain-exa langchain-brightdata langgraph langchainExa
Here, we’ve got a basic research agent. langchain_exa gives us access to ExaSearchResults and ExaFindSimilarResults. We begin by creating an instance of each one. Then, we set up a model using ChatOpenAI. We use LangGraph’s InMemorySaver class for basic agent memory. We then pass our tools and model into create_agent() to create our AI agent. Make note of our system prompt — we tell the agent to cite all relevant sources and give us a confidence score.
from langchain_openai import ChatOpenAI
from langchain_exa import ExaSearchResults, ExaFindSimilarResults
from langchain.agents import create_agent
from langgraph.checkpoint.memory import InMemorySaver
OPENAI_API_KEY = "your-openai-api-key"
EXA_API_KEY = "your-exa-api-key"
search_results = ExaSearchResults(
exa_api_key=EXA_API_KEY
)
similar_results = ExaFindSimilarResults(
exa_api_key=EXA_API_KEY
)
TOOLS = [search_results, similar_results]
model = ChatOpenAI(
model="gpt-5.4-mini",
temperature=0.1,
max_tokens=1000,
timeout=30,
api_key=OPENAI_API_KEY,
# ... (other params)
)
SYSTEM_PROMPT = """
You are a detailed research assistant.
Use the exa tools available to you in order to crawl and find latest global news.
Return a summary of each major news event with citations.
Give a confidence score for each story. If a story is corroborated by two sources, cite them both.
"""
agent = create_agent(model, tools=TOOLS, system_prompt=SYSTEM_PROMPT, checkpointer=InMemorySaver())
for chunk in agent.stream(
{"messages": [{"role": "user", "content": "Find the latest global news"}]},
{"configurable": {"thread_id": "1"}},
stream_mode="values"
):
msg = chunk["messages"][-1]
msg.pretty_print()Now, let’s run it and take a look at the output. Below is a snippet from LangChain’s “Tool Message.” As you can see, we get a title, URL, ID, score, date as well as some metadata and text from the site.
Exa tool message
Here’s the corresponding output generated by our AI agent. The AI agent ranked the story and gave a summary. It cited BBC live coverage as well as a BBC summary. It gave a confidence score of 0.93.
Exa agent output
Bright Data
Our Bright Data example is more or less the same. However, this agent uses Bright Data’s SERP API, which provides structured JSON results. We tell the agent to use the Unlocker API to verify stories.
from langchain_openai import ChatOpenAI
from langchain_brightdata import BrightDataSERP, BrightDataUnlocker
from langchain.agents import create_agent
from langgraph.checkpoint.memory import InMemorySaver
OPENAI_API_KEY = "your-openai-api-key"
BRIGHTDATA_API_KEY = "your-bright-data-api-key"
search_results = BrightDataSERP(
bright_data_api_key = BRIGHTDATA_API_KEY
)
similar_results = BrightDataUnlocker(
bright_data_api_key = BRIGHTDATA_API_KEY,
)
TOOLS = [search_results, similar_results]
model = ChatOpenAI(
model="gpt-5.4-mini",
temperature=0.1,
max_tokens=1000,
timeout=30,
api_key=OPENAI_API_KEY,
# ... (other params)
)
SYSTEM_PROMPT = """
You are a detailed research assistant.
Use the bright data tools available to you in order to crawl and find latest global news.
Return a summary of each major news event with citations. Verify each page using the unlocker tool.
Give a confidence score for each story. If a story is corroborated by two sources, cite them both.
"""
agent = create_agent(model, tools=TOOLS, system_prompt=SYSTEM_PROMPT, checkpointer=InMemorySaver())
for chunk in agent.stream(
{"messages": [{"role": "user", "content": "Find the latest global news"}]},
{"configurable": {"thread_id": "1"}},
stream_mode="values"
):
msg = chunk["messages"][-1]
msg.pretty_print()Now, let’s take a look at Bright Data’s tool outputs. Here’s some sample output from the SERP API. We get structured JSON results.
Bright Data SERP tool output
And here’s some sample output from the Unlocker API. The AI agent was able to access the full web page as markdown content.
Unlocker API sample output
Here’s one of the outputs from the AI agent. The agent actually cited the same story listed above by the Exa agent. The agent cited the story on three separate sources and gave a confidence score of 0.98.
Bright Data AI agent output
Conclusion
Exa and Bright Data give your AI agents web access that helps them discover news and trends in real time. AI agents can use unblocking tools like the Unlocker API to verify the news they find when searching. For efficient task management, AI agents need to be able to search and extract web data. For teams who need AI-native search and summary features, Exa offers a very strong product. If your team needs broader access to a full stack of data tools, Bright Data is better suited to fill those needs.
