Search returns pages; research returns a synthesised answer with sources, and it takes long enough that returning a task id is the only sane design. That fits an agent workflow well — kick off the investigation, do something else, collect the result. The two search depths matter for cost and latency: use standard for a lookup and deep when the answer is genuinely spread across sources.
Linkup's endpoint for real-time web search. The distinctive tool is the research one — a long autonomous investigation that returns a task id rather than an answer, with a cited synthesis when it finishes.
- Linkup-search — standard or deep, with domain filters, a date range, a result cap and optional images
- Linkup-research — an autonomous long-running investigation, returning a task id
- Linkup-get-research — poll that id for status and the cited answer
- Linkup-fetch — extract a page, optionally rendering JavaScript, with images or raw HTML
An API key from their app. A local npm package and a single-file Claude Desktop bundle are also published.
Paste a URL with a key — add the endpoint to your client