exa

Troubleshooting

If requests fail, run zero doctor check-connector --env-name EXA_TOKEN or zero doctor check-connector --url https://api.exa.ai/search --method POST

Official Docs

• Search API: https://exa.ai/docs/reference/search
• Contents API: https://exa.ai/docs/reference/contents
• Answer API: https://exa.ai/docs/reference/answer

Prerequisites

Connect the Exa connector at app.vm0.ai/connectors.

Search

Write to /tmp/exa_search.json:

{
  "query": "latest developments in LLM agents",
  "type": "auto",
  "contents": {
    "highlights": true
  }
}

Then run:

curl -s -X POST "https://api.exa.ai/search" --header "x-api-key: $EXA_TOKEN" --header "Content-Type: application/json" -d @/tmp/exa_search.json | jq .

Answer

Write to /tmp/exa_answer.json:

{
  "query": "What changed recently in AI agent infrastructure?",
  "text": true
}

Then run:

curl -s -X POST "https://api.exa.ai/answer" --header "x-api-key: $EXA_TOKEN" --header "Content-Type: application/json" -d @/tmp/exa_answer.json | jq .

Guidelines

• Use contents.text, contents.highlights, or contents.summary when the answer needs source material, not just URLs.
• Prefer type: "auto" unless the user specifically requests neural, fast, or deep search.
• Treat generated answers as grounded summaries and inspect citations before using them in high-stakes contexts.