> For the complete documentation index, see [llms.txt](https://docs.appmixer.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.appmixer.com/connector-configuration/exa.md).

# Exa

https\://dashboard.exa.ai

The Exa connector (*Search Web*, *Find Similar Links*, *Extract Page Contents*, *Answer Question*, *New Search Result*, *Make API Call*) authenticates with an Exa API key. Each user enters their own key when they connect an account in the Designer, so you do not need to register an app with Exa.

Exa is a web search API built for AI agents. *Search Web* runs a semantic or keyword web search, *Find Similar Links* returns pages similar to a given URL, *Extract Page Contents* turns a list of URLs into clean, LLM-ready text, *Answer Question* returns an AI-generated answer with the web sources it cited, *New Search Result* triggers when a search finds a page it has not returned before, and *Make API Call* reaches any other Exa endpoint at `https://api.exa.ai`.

## Get an API key

1. Sign up or sign in at [dashboard.exa.ai](https://dashboard.exa.ai). New accounts get free credits, no credit card needed.
2. Open the **API Keys** page and copy a key (or create a new one).
3. In the Appmixer Designer, connect an Exa account and paste the key into the *API Key* field.

Appmixer checks the key with an empty `POST https://api.exa.ai/contents` request. Exa rejects the empty body before it does any work, so the check uses no credits. Exa has no profile endpoint, so the connection is labelled with the masked key (for example `a1b2c3d4...wxyz`).

{% hint style="warning" %}
Treat the API key like a password. Anyone who has it can run searches that use the credits of your Exa account.
{% endhint %}

## Credits and limits

Exa bills every request against the credit balance of the account. See [Exa pricing](https://exa.ai/pricing) for the current prices. In short:

* A search costs more with a more thorough *Search Type*: *Instant* is the cheapest, *Fast* and *Auto* cost more, and the *Deep* modes cost the most.
* *Include Page Text*, *Include Highlights* and *Summary Query* are billed per returned page, on top of the search, and each one is billed separately.
* *Answer Question* is billed per request; *Include Citation Text* adds the page text price for every cited source.
* New accounts get free monthly credits.

Exa allows 10 requests per second on search and answer by default. The connector queues requests at 10 per second per user, so busy flows slow down instead of failing with rate limit errors.

## Searching

* **Query** is a natural-language description of what you want. Exa is semantic, so a descriptive phrase (`startups building AI agents for customer support`) works better than a list of keywords.
* **Search Type** picks the method: *Auto* lets Exa decide, *Fast* and *Instant* are quicker and cheaper, *Neural* and *Keyword* force one method, and the *Deep* modes search harder. Only *Neural* returns a relevance `score`.
* **Category** restricts the results to one kind of page, for example news, research papers, companies or GitHub.
* **Include Domains** / **Exclude Domains** take a comma-separated list of domains. **Published After** / **Published Before** filter by publication date.
* **Output Type** decides how the results leave the component: the whole list at once, the first result only, one result at a time, or a CSV file.
* When nothing matches, the component sends the input to its *Not Found* port instead (the query for *Search Web*, the URL for *Find Similar Links*, the URLs for *Extract Page Contents*), so the flow can branch on an empty result.

*Extract Page Contents* takes up to 100 URLs per call, separated by commas or new lines. **Max Cache Age (hours)** decides whether Exa may serve a cached copy (`0` forces a fresh crawl, `-1` serves cache only), and **Subpages** with **Subpage Target** also crawls linked pages such as `pricing` or `about`.

## New Search Result trigger

The trigger repeats the search on every poll and emits each page it has not returned before. The first poll only records the current results as a baseline, so pages that already exist when you start the flow do not fire it. New pages are detected among the top **Number Of Results** of each poll.

Every poll is a billed search. The trigger therefore offers only the cheap *Fast*, *Auto* and *Instant* search types, and *Fast* is the default.

## Make API Call

*Make API Call* takes a path relative to `https://api.exa.ai`, for example `/search`. A full URL is accepted only when it points at `https://api.exa.ai`. Any other host is refused, so the API key is never sent elsewhere. The key is added to every request in the `x-api-key` header. Most Exa data endpoints use `POST` with a JSON **Request Body**.

## Troubleshooting

* **"Invalid Exa API key."** when connecting: the key is wrong or was deleted. Copy it again from the **API Keys** page of the dashboard.
* **Credit or billing errors** from Exa: the account has run out of credits. Check the balance and billing in the Exa dashboard.
* **Nothing comes out of the&#x20;*****Out*****&#x20;port**: the search had no result, so the input went to the *Not Found* port. Connect it, or try a broader query, another category or fewer domain filters.
* **The trigger does not fire**: it fires only for pages that were not in the previous results. A query whose top results rarely change fires rarely; a higher *Number Of Results* widens the window.

## Backoffice configuration

No Backoffice configuration is required. The connector reads one optional key.

Configuration ID: `appmixer:ai:exa` (Backoffice → Configuration; see [Connector Configuration](https://docs.appmixer.com/getting-started/services) for how a configuration is created and how its keys reach the connector).

* `outputFilePrefix` – optional. File name prefix of the CSV file that *Search Web*, *Find Similar Links* and *Extract Page Contents* create when *Output Type* is set to a file; the file is named `<prefix>-<component name>.csv`. Default: `exa-objects-export`.

```bash
appmixer service-config set appmixer:ai:exa -u -d '{
  "outputFilePrefix": "exa-objects-export"
}'
```


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.appmixer.com/connector-configuration/exa.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
