DeepSeek Search and Fetch API vs Fetch (MCP reference server)

DeepSeek Search and Fetch API

Search returns up to 50 results with a model-length summary, an authority tier and optional full text, and only successful non-empty calls are billed, at $0.75 per 1,000. It opened on 17 October 2026, with 5 requests a second per account, no OpenAPI file, no free tier, and data held in the PRC under a policy that permits training.

Fetch (MCP reference server)

One tool of about 300 tokens, with max_length and start_index paging and a hint giving the next offset. Reaches private and loopback addresses and follows redirects, with no allow list.

DeepSeek Search and Fetch API

1. Keep query to 128 characters and put filters in it as operators, such as site:github.com lang:en after:2026-01-01
2. Leave need_content false and read summary first, then call Fetch on the URLs you keep
3. Stay under 5 requests a second per account across all keys. On 408, 429 or 5xx wait and retry, since no Retry-After is documented
4. Don't retry a Fetch 403 or 404 unchanged. The page is blocked by policy, unreachable or has no Markdown
5. Don't send private queries or an end user's IP in localization. Data is stored in the PRC

Fetch (MCP reference server)

1. Fetch with the 5,000-character default and continue from the start_index the truncation note gives
2. Expect empty bodies from JavaScript-rendered sites and switch to a browser tool
3. Treat a robots.txt refusal as final for autonomous calls and tell the user
4. Use raw: true only for non-HTML content. It skips the Markdown conversion and returns the page as served