Exa AI Search

Get API Key: https://exa.ai

LiteLLM Python SDK

Exa AI Search

import os
from litellm import search

os.environ["EXA_API_KEY"] = "exa-..."

response = search(
    query="latest AI developments",
    search_provider="exa_ai",
    max_results=5
)

LiteLLM AI Gateway

1. Setup config.yaml

config.yaml

model_list:
  - model_name: gpt-6.1-sol
    litellm_params:
      model: gpt-6.1-sol
      api_key: os.environ/OPENAI_API_KEY

search_tools:
  - search_tool_name: exa-search
    litellm_params:
      search_provider: exa_ai
      api_key: os.environ/EXA_API_KEY

2. Start the proxy

litellm --config /path/to/config.yaml

# RUNNING on http://0.0.0.0:4000

3. Test the search endpoint

Test Request

curl http://0.0.0.0:4000/v1/search/exa-search \
  -H "Authorization: Bearer $LITELLM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "latest AI developments",
    "max_results": 5
  }'

Provider-specific Parameters

Exa AI Search with Provider-specific Parameters

import os
from litellm import search

os.environ["EXA_API_KEY"] = "exa-..."

response = search(
    query="AI research papers",
    search_provider="exa_ai",
    max_results=10,
    search_domain_filter=["arxiv.org"],
    # Exa-specific parameters
    type="neural",                   # 'neural', 'keyword', or 'auto'
    contents={"text": True},         # Request text content
    use_autoprompt=True              # Enable Exa's autoprompt
)

Exa Response Fields

Exa results carry the same title, url, snippet and date fields as every other search provider. On top of those, each result keeps Exa's highlights, highlight_scores and score whenever Exa returned them, and the response keeps every top-level field Exa sent except results (such as requestId, searchTime and costDollars) under provider_metadata. A result Exa sent without these fields keeps the plain shape.

Field	Where	Present when
results[].highlights	each result	you request contents.highlights
results[].highlight_scores	each result	Exa scores the highlights
results[].score	each result	you use type="neural"
provider_metadata	the response	every Exa response

highlights are the sentences Exa picked from the page as the best match for your query. When you ask for highlights without text, snippet is those highlights joined together. provider_metadata.costDollars is what Exa charged for the call, so you can reconcile it with the cost LiteLLM tracked in x-litellm-response-cost and the spend logs.

Exa AI Search with highlights

import os
from litellm import search

os.environ["EXA_API_KEY"] = "exa-..."

response = search(
    query="regional bank net interest margin outlook",
    search_provider="exa_ai",
    max_results=2,
    type="neural",
    contents={"highlights": {"numSentences": 1}},
)

for result in response.results:
    print(result.score, result.highlights)

print(response.provider_metadata["costDollars"])

Same request through the AI Gateway

curl http://0.0.0.0:4000/v1/search/exa-search \
  -H "Authorization: Bearer $LITELLM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "regional bank net interest margin outlook",
    "max_results": 2,
    "type": "neural",
    "contents": {"highlights": {"numSentences": 1}}
  }'

Response

{
  "object": "search",
  "results": [
    {
      "title": "Q2 earnings: net interest margin up 1 basis point",
      "url": "https://example.com/q2-earnings",
      "snippet": "Net interest income grew $9.3 million linked quarter.",
      "date": "2026-08-31T15:00:25.000Z",
      "last_updated": null,
      "highlights": ["Net interest income grew $9.3 million linked quarter."],
      "score": 0.42
    }
  ],
  "provider_metadata": {
    "requestId": "1c8916cb62f8d5a7d6ea19515a26d5dc",
    "resolvedSearchType": "neural",
    "searchTime": 6.5,
    "costDollars": {"total": 0.004, "search": {"neural": 0.004}}
  }
}

Web search interception passes only each result's title, URL and snippet to the model, so these extra fields never reach the model's context.