Exa Search Integration Guide

Exa AI is a web_search provider offering neural, keyword, and hybrid search capabilities along with built-in content extraction for highlights, text, and summaries.

Install plugin

openclaw plugins install @openclaw/exa-plugin
openclaw gateway restart

Get an API key

Create an account

Create an account at exa.ai and obtain an API key through your dashboard.

Store the key

Configure EXA_API_KEY in the Gateway environment, or use the following setup:

openclaw configure --section web

Config

{
  plugins: {
    entries: {
      exa: {
        config: {
          webSearch: {
            apiKey: "exa-...", // optional if EXA_API_KEY is set
            baseUrl: "https://api.exa.ai", // optional; OpenClaw appends /search
          },
        },
      },
    },
  },
  tools: {
    web: {
      search: {
        provider: "exa",
      },
    },
  },
}

Environment alternative: place EXA_API_KEY in the Gateway environment. For a gateway installation, store it in ~/.openclaw/.env. Refer to Env vars.

Base URL override

Assign plugins.entries.exa.config.webSearch.baseUrl to direct Exa search requests through a compatible proxy or alternative endpoint. OpenClaw normalizes bare hosts by adding https:// at the beginning and /search at the end unless the path already terminates there. The final endpoint becomes part of the search cache key, so results from different endpoints are never combined.

Tool parameters

query (string, required), The search term.
count (number, default: 5), Number of results (1-100, limited by Exa search type).
type (auto' | 'neural' | 'fast' | 'deep' | 'deep-reasoning' | 'instant), The search type.
freshness (day' | 'week' | 'month' | 'year), Restricts results by time. Not usable with date_after/date_before.
date_after (string), Results after this date (YYYY-MM-DD).
date_before (string), Results before this date (YYYY-MM-DD).
contents (object), Controls for content extraction (details below).

Content extraction

Provide a contents object to manage what content is extracted from results:

await web_search({
  query: "transformer architecture explained",
  type: "neural",
  contents: {
    text: true, // full page text
    highlights: { numSentences: 3 }, // key sentences
    summary: true, // AI summary
  },
});

Contents option	Type	Description
`text`	`boolean \| { maxCharacters }`	Retrieve full page text
`highlights`	`boolean \| { maxCharacters, query, numSentences, highlightsPerUrl }`	Retrieve key sentences
`summary`	`boolean \| { query }`	AI-generated summary

When contents is not provided, Exa uses { highlights: true } as the default, meaning results include key-sentence excerpts. Descriptions for results are drawn from highlights first, then summary, then full text, whichever is available earliest. Results also retain the raw highlightScores and summary fields from the Exa API response when they exist.

Search modes

Mode	Description
`auto`	Exa selects the optimal mode (default)
`neural`	Search based on meaning and semantics
`fast`	Fast keyword-based search
`deep`	In-depth thorough search
`deep-reasoning`	Deep search with reasoning
`instant`	Quickest possible results

Notes

count supports up to 100 results, depending on Exa search type limits.
Results are cached for 15 minutes by default. Adjust the shared tools.web.search.cacheTtlMinutes (in minutes) and tools.web.search.timeoutSeconds (default 30s) to modify caching and request timeout for all web_search providers, including Exa.

Web Search overview -- all providers and automatic detection
Brave Search -- structured results with country and language filters
Perplexity Search -- structured results with domain filtering