Exa Get Contents

Exa Get Contents (Token-Efficient)

Use this skill to extract clean, LLM-ready content from known URLs using the Exa API via curl, while strictly minimizing token usage to keep context sizes manageable.

Note: This is for when you already have the URLs. If you need to search the web to find URLs first, use the exa-search skill instead.

Extraction Approaches

Choose the extraction mode that best balances your need for information against token cost. Do NOT wrap extraction options in a contents object (unlike the /search endpoint).

1. Highlights Extraction (Recommended)

Extracts key excerpts relevant to a query. This uses 10x fewer tokens than full text.

curl -X POST 'https://api.exa.ai/contents' \
  -H "x-api-key: $EXA_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{
  "urls": ["https://example.com/article", "https://example.com/another-post"],
  "highlights": {
    "query": "key findings and methodology",
    "maxCharacters": 4000
  }
}'

2. Structured Summary

Uses an LLM to generate an abstract or structured JSON. Excellent for analyzing dense company pages or long documents without reading the entire text.

curl -X POST 'https://api.exa.ai/contents' \
  -H "x-api-key: $EXA_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{
  "urls": ["https://example.com"],
  "summary": {
    "query": "Extract company overview and recent news",
    "schema": {
      "type": "object",
      "properties": {
        "name": { "type": "string" },
        "industry": { "type": "string" },
        "recent_news": { "type": "array", "items": { "type": "string" } }
      },
      "required": ["name", "industry"]
    }
  }
}'

3. Subpage Crawling (Documentation/Deep Dives)

Automatically discover and extract content from linked subpages (e.g., API docs). Use subpageTarget to focus the crawl.

curl -X POST 'https://api.exa.ai/contents' \
  -H "x-api-key: $EXA_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{
  "urls": ["https://docs.example.com"],
  "subpages": 5,
  "subpageTarget": ["api", "reference", "guide"],
  "highlights": { "maxCharacters": 4000 }
}'

Gotchas & Error Handling

• Top-Level Keys: text, highlights, and summary MUST be placed at the top level of the JSON payload. Do NOT nest them inside a contents object (that is only for the /search endpoint).
• Silent Failures: The API returns HTTP 200 even if individual URLs fail to scrape. You MUST check the statuses array in the JSON response to verify success.
• E.g., check that statuses[].status is "success" and not "error". Look for tags like CRAWL_NOT_FOUND or CRAWL_LIVECRAWL_TIMEOUT if an error occurs.
• Freshness: To bypass cache and force a live crawl, add "maxAgeHours": 0 and a timeout "livecrawlTimeout": 15000 to the top level of the JSON. Do not use the deprecated livecrawl parameter.
• Deprecated Parameters: Never use useAutoprompt, stream, tokensNum, numSentences, or highlightsPerUrl.
• Token warnings: Avoid text: true without limits. If you must use text, always cap it: "text": {"maxCharacters": 8000}.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: RowanAldean
• Source: RowanAldean/exa-skills
• License: Unlicense

Install and usage instructions live in the source repository linked above.