> ## Documentation Index
> Fetch the complete documentation index at: https://docs.agentruntime.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Exa connector

> Semantic web search, similar-page discovery, and cited answers via the Exa API.

The **exa** MCP adapter connects to [Exa](https://exa.ai/) for neural and keyword web search. Find relevant pages, discover content similar to a URL, and get AI-generated answers with citations — ideal for research workflows and grounded agent responses.

## Prerequisites

* An Exa account with an API key
* **project\_contributor** access in AgentRuntime

## Connect in AgentRuntime

<Steps>
  <Step title="Get an Exa API key">
    Sign up at [exa.ai](https://exa.ai/) and copy your API key from the dashboard.
  </Step>

  <Step title="Create a connection">
    Go to **Connections** and click **New custom connection** to create an API key connection with `APIKey`.
  </Step>

  <Step title="Add exa MCP instance">
    Go to **MCP**, click **Add instance** for **exa**, wire your connection on **Instance config**, set the profile **active**, and save.
  </Step>

  <Step title="Test search">
    Run **mcp\_call** → `exa_search` with a simple query to confirm connectivity.
  </Step>
</Steps>

See [Authentication](/integrations/authentication) for binding connections to MCP instances.

## What you can build

* **Research briefs** — `exa_search` with `include_text=true` → LLM synthesizes findings → save to Notion or Google Docs.
* **Competitive intel** — `exa_find_similar` on a competitor landing page → rank overlapping products.
* **Q\&A with citations** — `exa_answer` for user questions in Chat Autopilot → display answer plus source URLs.
* **Deep research mode** — `exa_search` with `type=deep` for multi-hop investigation before human review.

## Tools

| Tool               | Description                                                              |
| ------------------ | ------------------------------------------------------------------------ |
| `exa_search`       | Web search with modes `auto`, `fast`, or `deep`; optional full page text |
| `exa_find_similar` | Find pages semantically similar to a given URL                           |
| `exa_answer`       | AI-generated answer with cited web sources                               |

## Example

Search the web and pass results to an LLM step:

```json theme={null}
{
  "id": "exa-research",
  "type": "mcp_call",
  "name": "Research topic",
  "tool_name": "exa_search",
  "tool_args": {
    "query": "{{input.topic}} market trends 2026",
    "num_results": 8,
    "type": "auto",
    "include_text": true
  },
  "timeout_s": 45
}
```

## Configuration

| Key       | Required | Description      |
| --------- | -------- | ---------------- |
| `APIKey`  | Yes      | Exa API key      |
| `APIBase` | No       | Exa API base URL |

## Troubleshooting

| Issue                           | Fix                                                           |
| ------------------------------- | ------------------------------------------------------------- |
| Empty results                   | Broaden query; try `type=deep` for niche topics               |
| Rate limits                     | Reduce `num_results` or add backoff between workflow runs     |
| Stale citations in `exa_answer` | Re-run search; Exa synthesizes from live index at query time  |
| High latency on deep search     | Increase step `timeout_s`; deep mode takes longer than `fast` |

## Related

* [Integrations quickstart](/integrations/quickstart)
* [Brave connector](/connectors/brave) — alternative web search
* [Connector catalog](/integrations/connector-catalog)
* [Troubleshooting](/platform/troubleshooting)