Web Search API (Exa)

Neural web search engine designed for AI agents. Semantic ranking via neural embeddings (not just keywords), similar page discovery for competitive research, full-text page content extraction, and AI-grounded answers with source citations. Filter by category: company, research paper, news, PDF, GitHub, LinkedIn profile, financial report.

Base URL: https://api.jarvisclaw.ai/v1/marketplace/exa

Authentication

Both methods are supported — all requests settle via x402 on-chain:

Method	Header	Description
API Key	`Authorization: Bearer sk-...`	Platform signs x402 from your HD wallet automatically
Private Key (x402)	Automatic via SDK	Agent signs x402 directly from its own wallet

See Agent Payments (x402) for full details on how both methods work.

Endpoints

Method	Endpoint	Description
POST	`/search`	Semantic web search with neural ranking
POST	`/find-similar`	Find pages similar to a given URL
POST	`/contents`	Extract full-text content from URLs
POST	`/answer`	AI-grounded answer with source citations

Pricing

Reference rates per request. For the exact charge on a given call, read price.amount in the response.

Endpoint	Price	Description
`/search`	$0.012/request	Neural semantic search
`/find-similar`	$0.012/request	Similar page discovery (not yet available)
`/contents`	$0.004/request	Full-text page extraction
`/answer`	$0.012/request	AI answer with citations
`/highlights`	$0.002/request	Relevant passage extraction

POST /search

Search the web using neural semantic ranking. Returns results ranked by meaning, not just keyword matching.

Parameters

Parameter	Type	Required	Description
`query`	string	Yes	Search query (max 10,000 characters)
`numResults`	integer	No	Number of results to return. Default: `10`
`category`	string	No	Filter by category: `company`, `research paper`, `news`, `pdf`, `github`, `linkedin profile`, `financial report`
`startPublishedDate`	string	No	Filter results published after this date (ISO 8601)
`endPublishedDate`	string	No	Filter results published before this date (ISO 8601)

Request

json

{
  "query": "latest breakthroughs in protein folding AI models 2026",
  "numResults": 5,
  "category": "research paper",
  "startPublishedDate": "2026-01-01T00:00:00Z"
}

Response

json

{
  "results": [
    {
      "title": "AlphaFold 3: Predicting All Molecular Interactions",
      "url": "https://example.com/alphafold3-paper",
      "score": 0.95,
      "publishedDate": "2026-03-15T00:00:00Z",
      "author": "DeepMind Research"
    },
    {
      "title": "ESMFold Advances in Zero-Shot Structure Prediction",
      "url": "https://example.com/esmfold-advances",
      "score": 0.89,
      "publishedDate": "2026-02-20T00:00:00Z",
      "author": "Meta AI"
    }
  ],
  "query": "latest breakthroughs in protein folding AI models 2026"
}

POST /find-similar Coming Soon

Find web pages similar to a given URL. Useful for competitive research, finding related content, and discovering alternatives.

WARNING

This endpoint is not yet available. It will return 404 until enabled. Use /search with a descriptive query as an alternative.

Parameters

Parameter	Type	Required	Description
`url`	string	Yes	Source URL to find similar pages for (must be publicly indexable)
`numResults`	integer	No	Number of similar results to return. Default: `10`

Request

json

{
  "url": "https://openai.com/research/gpt-4",
  "numResults": 5
}

Response

json

{
  "results": [
    {
      "title": "Claude 3.5 Sonnet: Technical Report",
      "url": "https://example.com/claude-report",
      "score": 0.92
    },
    {
      "title": "Gemini Ultra: Capabilities and Limitations",
      "url": "https://example.com/gemini-ultra",
      "score": 0.87
    }
  ],
  "sourceUrl": "https://openai.com/research/gpt-4"
}

POST /contents

Extract full-text content from one or more URLs. Returns clean, parsed page text suitable for LLM consumption.

Parameters

Parameter	Type	Required	Description
`ids`	string[]	Yes	Array of URLs to extract content from

The field is ids, not urls

Exa names this field ids even though it carries URLs. Both SDKs send ids; sending urls returns no content.

Request

json

{
  "ids": [
    "https://example.com/article-1",
    "https://example.com/article-2"
  ]
}

Response

json

{
  "results": [
    {
      "id": "https://example.com/article-1",
      "url": "https://example.com/article-1",
      "title": "Understanding Transformer Architectures",
      "text": "Transformer models have revolutionized natural language processing...",
      "publishedDate": "2026-05-10T00:00:00Z"
    },
    {
      "id": "https://example.com/article-2",
      "url": "https://example.com/article-2",
      "title": "Scaling Laws for Neural Language Models",
      "text": "We study empirical scaling laws for language model performance...",
      "publishedDate": "2026-04-22T00:00:00Z"
    }
  ]
}

Results come back under results. The Python SDK's contents() reads results (falling back to data), so treat contents as absent.

POST /answer

Get an AI-generated answer to a query, grounded in web sources with citations.

Parameters

Parameter	Type	Required	Description
`query`	string	Yes	Question to answer (max 10,000 characters)

Request

json

{
  "query": "What are the main differences between GPT-4 and Claude 3.5?"
}

Response

json

{
  "answer": "GPT-4 and Claude 3.5 differ in several key areas: architecture design, context window size, multimodal capabilities, and safety approaches. GPT-4 supports up to 128k tokens of context while Claude 3.5 Sonnet supports 200k tokens...",
  "sources": [
    {
      "title": "GPT-4 Technical Report",
      "url": "https://example.com/gpt4-report"
    },
    {
      "title": "Claude 3.5 Model Card",
      "url": "https://example.com/claude-model-card"
    }
  ]
}

Errors

HTTP Status	Code	Description
400	`query_too_long`	Query exceeds the 10,000 character limit
400	`invalid_url`	The provided URL is malformed or not a valid web address
422	`url_unreachable`	The URL could not be fetched (not publicly accessible or blocked)

Code Examples

bash

# Semantic search
curl -X POST https://api.jarvisclaw.ai/v1/marketplace/exa/search \
  -H "Authorization: Bearer sk-your-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "best open source LLM frameworks 2026",
    "numResults": 5,
    "category": "github"
  }'

# Find similar pages
curl -X POST https://api.jarvisclaw.ai/v1/marketplace/exa/find-similar \
  -H "Authorization: Bearer sk-your-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://github.com/langchain-ai/langchain",
    "numResults": 5
  }'

# Extract page contents (field is "ids", not "urls")
curl -X POST https://api.jarvisclaw.ai/v1/marketplace/exa/contents \
  -H "Authorization: Bearer sk-your-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "ids": ["https://example.com/article"]
  }'

# Get AI answer with citations
curl -X POST https://api.jarvisclaw.ai/v1/marketplace/exa/answer \
  -H "Authorization: Bearer sk-your-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "What is retrieval augmented generation and how does it work?"
  }'

python

import requests

BASE = "https://api.jarvisclaw.ai/v1/marketplace/exa"
HEADERS = {
    "Authorization": "Bearer sk-your-api-key",
    "Content-Type": "application/json",
}

# Semantic search
resp = requests.post(f"{BASE}/search", headers=HEADERS, json={
    "query": "best open source LLM frameworks 2026",
    "numResults": 5,
    "category": "github",
})
for result in resp.json()["results"]:
    print(f"{result['title']}: {result['url']}")

# Find similar pages (coming soon — endpoint not yet available)
# resp = requests.post(f"{BASE}/find-similar", headers=HEADERS, json={
#     "url": "https://github.com/langchain-ai/langchain",
#     "numResults": 5,
# })
# for result in resp.json()["results"]:
#     print(f"{result['title']} (score: {result['score']})")

# Extract page contents (field is "ids", not "urls")
resp = requests.post(f"{BASE}/contents", headers=HEADERS, json={
    "ids": ["https://example.com/article-1", "https://example.com/article-2"],
})
for content in resp.json().get("results", []):
    print(f"{content.get('title')}: {content.get('text', '')[:100]}...")

# AI-grounded answer
resp = requests.post(f"{BASE}/answer", headers=HEADERS, json={
    "query": "What is retrieval augmented generation and how does it work?",
})
data = resp.json()
print(data["answer"])
for source in data["sources"]:
    print(f"  Source: {source['url']}")

python

from jarvisclaw import SearchClient

# --- Option A: Base chain (EVM) ---
# Hex private key -> USDC on Base (Chain ID 8453)
search = SearchClient(private_key="0x<evm-private-key>")

# --- Option B: Solana ---
# Base58 keypair -> USDC SPL on Solana mainnet
# search = SearchClient(private_key="<solana-bs58-keypair>")

# SDK auto-detects chain from key format - no config needed

# NOTE: query() posts to /v1/search (smart-routed summary), NOT to the Exa
# endpoint above. It usually returns a single summary result with url="".
results = search.query("best open source LLM frameworks 2026", num_results=5)
for r in results:
    print(f"{r.title}: {r.url}")

# For per-result titles and URLs, call the Exa endpoint directly:
import requests
resp = requests.post(
    "https://api.jarvisclaw.ai/v1/marketplace/exa/search",
    headers={"Content-Type": "application/json"},
    json={"query": "best open source LLM frameworks 2026", "numResults": 5},
)

# Find similar pages (coming soon — not yet available)
# similar = search.find_similar("https://github.com/langchain-ai/langchain", num_results=5)
# for r in similar:
#     print(f"{r.title}: {r.url}")

# Extract page contents
contents = search.contents(["https://example.com/article-1", "https://example.com/article-2"])
for c in contents:
    print(c)

package main

import (
    "context"
    "fmt"
    jc "github.com/api-jarvisclaw/go-sdk/v2"
)

func main() {
    ctx := context.Background()
    sc, _ := jc.NewSearchClient(jc.WithAPIKey("sk-your-api-key"))

    // Semantic search — returns *SearchResponse
    resp, _ := sc.Query(ctx, "best open source LLM frameworks 2026",
        jc.WithNumResults(5))
    fmt.Printf("Summary: %s\n", resp.Summary)
    fmt.Printf("Sources used: %d\n", resp.SourcesUsed)
    for _, r := range resp.Citations {
        fmt.Printf("  %s: %s\n", r.Title, r.URL)
    }
}

package main

import (
    "context"
    "fmt"
    jc "github.com/api-jarvisclaw/go-sdk/v2"
)

func main() {
    ctx := context.Background()

    // x402 Agent wallet - pays per-call via USDC on Base (Chain ID 8453)
    sc, _ := jc.NewSearchClient(jc.WithPrivateKey("0x<evm-private-key>"))

    // Semantic search — returns *SearchResponse
    resp, _ := sc.Query(ctx, "best open source LLM frameworks 2026",
        jc.WithNumResults(5))
    fmt.Printf("Summary: %s\n", resp.Summary)
    fmt.Printf("Sources used: %d\n", resp.SourcesUsed)
    for _, r := range resp.Citations {
        fmt.Printf("  %s: %s\n", r.Title, r.URL)
    }
}

Two different search surfaces

This page documents the Exa marketplace endpoints under /v1/marketplace/exa/*, which return structured results with individual titles, URLs, and scores.

There is also POST /v1/search — the smart-routed surface behind model: auto/search. It returns an AI-written summary with inline citations, not a ranked list:

json

{
  "query": "...",
  "summary": "Retrieval Augmented Generation is...[1](https://example.com)...",
  "citations": [],
  "sources_used": 10
}

Need	Use
Ranked links with URLs and scores	`POST /v1/marketplace/exa/search`
A written answer with citations	`POST /v1/search`, or `/v1/marketplace/exa/answer`

The SDK helpers split along the same line: SearchClient.query() hits /v1/search, while SearchClient.contents() and find_similar() hit the Exa endpoints. See Smart Router → auto/search.

Limitations

10,000 character query max — queries exceeding this length are rejected with query_too_long
Public pages only — cannot access content behind authentication or paywalls
Concurrency cap — heavy parallel use can return 429; back off and retry
find-similar is not yet available — the upstream provider has not registered it; it returns 404
answer takes 3-8 seconds — the AI synthesis step adds latency compared to raw search

Web Search API (Exa) ​

Authentication ​

Endpoints ​

Pricing ​

POST /search ​

Parameters ​

Request ​

Response ​

POST /find-similar Coming Soon ​

Parameters ​

Request ​

Response ​

POST /contents ​

Parameters ​

Request ​

Response ​

POST /answer ​

Parameters ​

Request ​

Response ​

Errors ​

Code Examples ​

Two different search surfaces ​

Limitations ​

Web Search API (Exa)

Authentication

Endpoints

Pricing

POST /search

Parameters

Request

Response

POST /find-similar Coming Soon

Parameters

Request

Response

POST /contents

Parameters

Request

Response

POST /answer

Parameters

Request

Response

Errors

Code Examples

Two different search surfaces

Limitations