Exa

Connect to Exa for AI-powered semantic web search, content enrichment, finding similar pages, website crawling, direct answers, structured research, and large-scale URL discovery.

Supports authentication: API Key

Set up the agent connector

Register your Exa API key with Scalekit so it can authenticate and proxy requests on behalf of your users. Unlike OAuth connectors, Exa uses API key authentication — there is no redirect URI or OAuth flow.

1. Generate an Exa API key

   • Sign in to dashboard.exa.ai/api-keys. Under Management, click API Keys.

   • Click + Create Key, enter a name (e.g., Agent Auth), and confirm.

   • In the Secret Key column, click the eye icon to reveal the key and copy it. Store it somewhere safe — you will not be able to view it again.

   

2. Create a connection in Scalekit

   • In Scalekit dashboard, go to Agent Auth → Create Connection. Find Exa and click Create.

   • Note the Connection name — you will use this as connection_name in your code (e.g., exa).

   

3. Add a connected account

   Connected accounts link a specific user identifier in your system to an Exa API key. You can add them via the dashboard for testing, or via the Scalekit API in production.

   Via dashboard (for testing)

   • Open the connection you created and click the Connected Accounts tab → Add account.

   • Fill in:

     • Your User’s ID — a unique identifier for this user in your system (e.g., user_123)
     • API Key — the Exa API key you copied in step 1

   • Click Save.

   

   Via API (for production)

   Python

   scalekit_client.actions.upsert_connected_account(
       connection_name="exa",
       identifier="user_123",       # your user's unique ID
       credentials={"api_key": "your-exa-api-key"}
   )

   Tip

   In production, call upsert_connected_account when a user connects their Exa account — for example, after they paste their API key into a settings page in your app.

Note

Credits and rate limits: Each Exa API key has a default limit of 10 QPS. Search, find-similar, and get-contents cost 1 credit per request, plus additional credits per content item (text, highlights, or summary) returned. exa_research and exa_websets run multiple sub-queries internally and consume significantly more credits. Monitor usage at dashboard.exa.ai → Usage.

Usage

Once a connected account is set up, make API calls through the Scalekit proxy. Scalekit injects the Exa API key automatically — you never handle credentials in your application code.

Python

import scalekit.client, os
from dotenv import load_dotenv
load_dotenv()

connection_name = "exa"    # connection name from your Scalekit dashboard
identifier = "user_123"    # your user's unique identifier

# Get your credentials from app.scalekit.com → Developers → Settings → API Credentials
scalekit_client = scalekit.client.ScalekitClient(
    client_id=os.getenv("SCALEKIT_CLIENT_ID"),
    client_secret=os.getenv("SCALEKIT_CLIENT_SECRET"),
    env_url=os.getenv("SCALEKIT_ENV_URL"),
)
actions = scalekit_client.actions

# Make a request via Scalekit proxy — no API key needed here
result = actions.request(
    connection_name=connection_name,
    identifier=identifier,
    path="/search",
    method="POST",
    json={"query": "LLM observability tools 2025", "num_results": 5}
)
print(result)

Tip

Exa uses API key auth — unlike OAuth connectors, there is no authorization link or redirect flow. Once you call upsert_connected_account (or add an account via the dashboard), your users can make requests immediately.

Tool list

exa_search

Search the web with a semantic or keyword query. Returns ranked results with titles, URLs, published dates, and optional full-page content. Semantic search (type: neural) finds conceptually similar pages even when your exact words don’t appear in the source. Keyword search (type: keyword) uses traditional exact-match scoring.

Credits: 1 credit per request + 1 credit per result when contents.text or contents.summary is requested.

Name	Type	Required	Description
query	string	Yes	Search query — phrase it as a statement or question for best semantic results
num_results	integer	No	Number of results to return (default 10, max 100)
type	string	No	Search mode: neural for semantic similarity (default), keyword for exact-match
category	string	No	Filter by content type: company, research paper, news, github, tweet, personal site, pdf, linkedin profile
include_domains	array	No	Restrict results to these domains (e.g., ["techcrunch.com", "arxiv.org"])
exclude_domains	array	No	Exclude results from these domains
start_published_date	string	No	ISO 8601 date — only return pages published after this date (e.g., 2025-01-01)
end_published_date	string	No	ISO 8601 date — only return pages published before this date
start_crawl_date	string	No	ISO 8601 date — only return pages crawled by Exa after this date
end_crawl_date	string	No	ISO 8601 date — only return pages crawled by Exa before this date
include_text	array	No	Strings that must appear in the page text
exclude_text	array	No	Strings that must not appear in the page text
location	string	No	ISO country code for location-biased results (e.g., US, GB, DE)
contents.text	boolean	No	Include full page text in each result (costs 1 extra credit per result)
contents.highlights	object	No	Include highlighted snippets — set num_sentences and highlights_per_url
contents.summary	object	No	Include an AI-generated page summary — optionally set query to focus the summary

exa_find_similar

Find web pages that are semantically similar to a given URL. Exa uses its neural index to surface pages with matching content, tone, and topic — useful for competitive research, discovering similar products, or finding alternative sources on a topic.

Credits: 1 credit per request + 1 credit per result when contents.text or contents.summary is requested.

Name	Type	Required	Description
url	string	Yes	URL to find similar pages for (e.g., https://example.com/product)
num_results	integer	No	Number of results to return (default 10, max 100)
include_domains	array	No	Restrict results to these domains (e.g., ["techcrunch.com", "wired.com"])
exclude_domains	array	No	Exclude results from these domains
start_published_date	string	No	ISO 8601 date — only return pages published after this date
end_published_date	string	No	ISO 8601 date — only return pages published before this date
start_crawl_date	string	No	ISO 8601 date — only return pages crawled by Exa after this date
end_crawl_date	string	No	ISO 8601 date — only return pages crawled by Exa before this date
contents.text	boolean	No	Include full page text in each result (costs 1 extra credit per result)
contents.highlights	object	No	Include highlighted snippets — set num_sentences and highlights_per_url
contents.summary	object	No	Include an AI-generated page summary — optionally set query to focus the summary

exa_crawl

Crawl all pages of a website starting from a given URL and extract their text content. Follows internal links up to the configured depth. Use this to index documentation sites, company knowledge bases, or competitor content for downstream processing.

Credits: Credits vary based on the number of pages crawled and content extracted.

Name	Type	Required	Description
url	string	Yes	Starting URL to crawl (e.g., https://docs.example.com)
max_depth	integer	No	Maximum number of link hops to follow from the starting URL (default 1)
max_pages	integer	No	Maximum number of pages to retrieve (default 100) — set this to control credit usage
include_subdomains	boolean	No	Whether to follow links to subdomains of the starting URL (default false)
exclude_paths	array	No	URL path patterns to skip (e.g., ["/archive/*", "/blog/*"])

exa_get_contents

Retrieve enriched content — full text, highlighted snippets, or AI-generated summaries — for one or more specific URLs without running a search. Use this when you already have URLs (from a prior search, a CRM, or an external list) and want to extract structured content from them.

Credits: 1 credit per URL + 1 credit per content item retrieved (text, highlights, or summary each count separately).

Name	Type	Required	Description
urls	array	Yes*	List of URLs to retrieve content for (e.g., ["https://example.com/about"])
ids	array	Yes*	List of Exa result IDs from a prior exa_search or exa_find_similar call — use instead of urls when you have Exa IDs
text	boolean	No	Include full page text for each URL (costs 1 extra credit per page)
text.max_characters	integer	No	Truncate page text to this many characters
text.include_html_tags	boolean	No	Preserve HTML tags in the returned text — helps LLMs interpret page structure
highlights	object	No	Extract relevant text snippets from each page
highlights.query	string	No	Custom query to guide which snippets are highlighted
summary	object	No	Generate an AI summary for each page (costs 1 extra credit per page)
summary.query	string	No	Focus the summary on a specific question or topic
summary.schema	object	No	JSON schema for structured summary output
subpages	integer	No	Number of subpages to also crawl from each URL (default 0)
subpage_target	string	No	Keyword or path pattern to target specific subpages (e.g., "pricing")
max_age_hours	integer	No	Maximum age of cached content in hours — set a lower value to force a fresher crawl

*Provide either urls or ids — at least one is required.

exa_answer

Get a concise natural language answer to a question, synthesized from live web search results. Returns the answer text and a list of source URLs used to generate it. Use this when you need a direct factual response rather than a list of links.

Credits: Credits vary based on the number of sources retrieved.

Name	Type	Required	Description
query	string	Yes	The question to answer (e.g., "What are the rate limits for GPT-4o?")
num_results	integer	No	Number of web sources used to generate the answer (default 5)
text	boolean	No	Include the source text snippets alongside the answer
include_domains	array	No	Restrict source lookup to these domains
exclude_domains	array	No	Exclude these domains from the source lookup
start_published_date	string	No	ISO 8601 date — only use sources published after this date
end_published_date	string	No	ISO 8601 date — only use sources published before this date

exa_research

Run in-depth, multi-angle research on a topic. Exa decomposes the topic into multiple sub-queries, runs each in parallel, and synthesizes a structured output. Optionally provide a JSON schema to control the output shape. Use this for competitive intelligence, market research, or any question that requires aggregating information across many sources.

Credits: Significantly more expensive than exa_search — cost scales with num_subqueries. Each sub-query consumes search and content retrieval credits.

Name	Type	Required	Description
topic	string	Yes	Research topic or question (e.g., "Competitive landscape of AI coding assistants in 2025")
num_subqueries	integer	No	Number of parallel sub-queries to run (default 5, max 15) — higher values improve coverage but increase cost
output_schema	object	No	JSON schema defining the structure of the output — omit for free-form markdown output
include_domains	array	No	Restrict all sub-query sources to these domains

exa_websets

Execute a complex natural language query designed to return a large set of matching URLs — potentially thousands. Each result is vetted against your criteria before being returned. Use this for bulk data collection tasks such as building lead lists, aggregating industry news, or sourcing research datasets.

Credits: Credits are consumed per vetted result. Cost scales with count.

Name	Type	Required	Description
query	string	Yes	Natural language description of the set to find (e.g., "YC-backed AI startups founded after 2022")
count	integer	No	Target number of matching URLs to return (default 10, can be thousands)
entity_type	string	No	Type of entity to find: company, person, article, research paper, job listing, event, product
criteria	array	No	Additional filter criteria — each item has a description and options.type (must, should, or must_not)
include_domains	array	No	Restrict results to these domains (e.g., ["linkedin.com", "crunchbase.com"])