AI & model APIs · active

DeepSeek Search & Fetch

Dedicated web retrieval APIs for agent workflows, with tool-specific authentication, usage accounting and limits separate from model inference.

Deliberately narrow scope

This entry describes dedicated retrieval, not every DeepSeek model or client. The official Harness module distinguishes this path from search performed through a full Messages model turn. Do not use a tool-request price as the total cost of the latter workflow.

Integration boundary

Use the proxy credential and search 401 case to separate the tool route from chat configuration. Keep keys out of browser bundles and logs, and follow target websites’ access policies. Source review on October 27, 2026 did not include a live request, paid call or account setup.

Integration surface

Versioning
This profile covers the dedicated /v1/tools/search and /v1/tools/fetch routes, not the full model API.
Resource groups
Web search, Page fetch
Pagination
Search exposes a count parameter; cursor pagination was not established in this review.
Errors
Distinguish authentication 401, insufficient balance 402 and throttling 429.
Webhooks
No webhook contract was verified for these request-response tools.
Official SDKs
None confirmed in the reviewed official sources.

Practical integration cases