Search the web with your Concurred API key. Results come from Exa, which is built for AI-style natural-language queries.

POST /api/v1/search

When should I use this?

Use /api/v1/search when you want a plain list of search results that you'll handle yourself — to show as cards on a page, save to a database, or feed into your own prompt.

If instead you want an AI model to do the searching for you and give you an answer, use the Chat API's battle or fight modes — the models will search the web on their own when they need to.

Request

{
  "query": "latest AI research on mixture-of-experts",
  "num_results": 10,
  "type": "neural",
  "include_summary": true,
  "recency": "1m"
}

Field	Type	Required	Description
`query`	string	Yes	The search query. Max 1000 chars.
`num_results`	integer	No	How many results to return (1–25, default 10).
`type`	string	No	`"neural"` (default, semantic), `"keyword"` (lexical, cheaper), or `"auto"`.
`include_summary`	boolean	No	Include Exa's synthesized summary per result. Default `true`.
`recency`	string	No	Filter by publication recency. Format: `"7d"`, `"1m"`, `"1y"`, etc.

Response

{
  "success": true,
  "data": {
    "query": "latest AI research on mixture-of-experts",
    "results": [
      {
        "url": "https://arxiv.org/abs/2026.XXXXX",
        "title": "Scaling Sparse MoE Transformers...",
        "snippet": "We present a new routing algorithm...",
        "score": 0.94,
        "published_date": "2026-03-14",
        "author": "Smith et al.",
        "favicon": "https://arxiv.org/favicon.ico"
      }
    ],
    "auto_prompt": "Research papers on mixture-of-experts architectures from 2026",
    "cost_usd": 0.005,
    "request_id": "req_..."
  }
}

Field	Description
`results[].url`	Canonical URL of the result.
`results[].title`	Page title.
`results[].snippet`	Short text excerpt (up to 500 chars).
`results[].score`	Relevance score, 0–1.
`results[].published_date`	ISO date if Exa could detect one.
`auto_prompt`	Exa's rewritten version of your query (what it actually searched).
`cost_usd`	Per-request cost from Exa, surfaced for debugging.

Authentication

Same as the rest of the v1 gateway — Bearer token or X-API-Key:

curl -X POST https://concurred.ai/api/v1/search \
  -H "Authorization: Bearer ck_..." \
  -H "Content-Type: application/json" \
  -d '{"query": "latest AI research on mixture-of-experts"}'

Code examples

curl -X POST https://concurred.ai/api/v1/search \
  -H "Authorization: Bearer $CONCURRED_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "openai news today",
    "num_results": 5,
    "recency": "7d"
  }'

Errors

Status	Code	Meaning
400	`BAD_REQUEST`	Missing or invalid `query`.
401	`UNAUTHORIZED` / `INVALID_API_KEY`	Missing or bad API key.
429	`RATE_LIMIT_EXCEEDED`	Either Concurred's or Exa's rate limit. Retry after a few seconds.
500	`INTERNAL_ERROR`	`EXA_API_KEY` not configured on the deployment.
503	`SERVICE_UNAVAILABLE`	Exa returned a non-2xx.

Search API

Request

Response

Authentication

Code examples

Errors

On this page