Search & Fetch | DeepSeeokay API Docs

The Search API and the Fetch API give brokers and RAG purposes entry to dwell net content material:

• Search API finds net pages for a question and returns structured outcomes. It is powered by DeepSeeokay’s in-house search engine.
• Fetch API retrieves a single net web page and returns its content material as Markdown.

Both are uncooked JSON APIs. They use the identical bearer authentication because the DeepSeeokay API.

The base URL is https://api.deepseek.com:

See the Search API Reference and Fetch API Reference for the entire request and response schemas.

Authentication​

Send your API key and the JSON content material sort within the request headers:

Search API​

Request​

curl https://api.deepseek.com/v1/tools/search 
  -H "Authorization: Bearer " 
  -H "Content-Type: utility/json" 
  -d '{
    "question": "DeepSeeokay V4 launch",
    "depend": 5,
    "need_content": true
  }'

import requests

response = requests.publish(
    "https://api.deepseek.com/v1/tools/search",
    headers={
        "Authorization": "Bearer ",
    },
    json={
        "question": "DeepSeeokay V4 launch",
        "depend": 5,
        "need_content": True,
    },
)
response.raise_for_status()

for outcome in response.json()["results"]:
    print(outcome["title"], outcome["url"])
    print(outcome["summary"])

Request fields:

Unknown fields within the request physique are ignored.

Query Operators​

The question is forwarded with its search syntax intact. The examples beneath mix operators with a search time period; you possibly can mix a number of operators in a single question.

For instance:

{
  "question": "DeepSeeokay website:github.com OR website:huggingface.co lang:en after:2025",
  "depend": 10
}

Response​

A profitable request returns HTTP 200 with a outcomes array. A traditional search with no matches returns an empty array.

{
  "outcomes": [
    {
      "rerank_score": 2.4,
      "title": "DeepSeek V4 model released",
      "url": "https://example.com/deepseek-v4",
      "snippet": "DeepSeek V4 is now available, with tool calling and long-context support.",
      "summary": "DeepSeek has released the V4 model. The release adds improved tool calling and long-context support for agent workflows.",
      "content": "DeepSeek V4 is now generally available...",
      "site_name": "Example News",
      "site_icon": "https://example.com/favicon.ico",
      "date": "2026-06-01T00:00:00+00:00",
      "authority": 3
    }
  ]
}

Each merchandise in outcomes comprises:

Fetch API​

Request​

curl https://api.deepseek.com/v1/tools/fetch 
  -H "Authorization: Bearer " 
  -H "Content-Type: utility/json" 
  -d '{
    "url": "https://example.com/article"
  }'

import requests

response = requests.publish(
    "https://api.deepseek.com/v1/tools/fetch",
    headers={
        "Authorization": "Bearer ",
    },
    json={"url": "https://example.com/article"},
)
response.raise_for_status()

web page = response.json()
print(web page["title"])
print(web page["content"])

Request fields:

Response​

{
  "title": "Example article",
  "content material": "# Example articlennArticle content material."
}

Fetching is topic to the goal website’s availability and entry insurance policies.

Errors comply with the OpenAI-compatible error format. The param area is all the time null.

{
  "error": {
    "message": "'depend' have to be between 1 and 50",
    "sort": "invalid_request_error",
    "param": null,
    "code": "invalid_request_error"
  }
}

Search and Fetch are read-only. For HTTP 408, 429, or 5xx errors, wait and retry. For HTTP 402, high up your account earlier than retrying. Do not retry different 4xx errors with out altering the request.

Billing​

Search and Fetch are billed per profitable request, not by tokens or by the variety of returned outcomes:

A request is billable solely when it returns HTTP 200 and produces a outcome (Search should return a non-empty outcomes array, and Fetch should return web page content material). Failed requests, empty search outcomes, and timeouts aren’t billed. depend solely controls what number of outcomes are returned and doesn’t have an effect on the per-request worth.

Note: whenever you use the DeepSeeokay API by means of DeepSeeokay Harness, Claude Code, or Codex, server-side instrument calls already hit the Search and Fetch APIs. These calls are at the moment freed from cost; per-request billing begins at 00:00 Beijing time on November 1, 2026.

Rate Limits​

For every account, Search and Fetch are restricted independently, and visitors to 1 doesn’t eat the opposite’s quota:

The QPS (requests per second) limits are calculated on the account degree, no matter which API key’s used, and are separate from the mannequin concurrency limits. Requests above the account-level QPS restrict obtain HTTP 429. In addition to the account-level restrict, Fetch can be topic to a per-domain QPS restrict, and requests above that restrict additionally obtain HTTP 429.