# LinkupPlatform-linkup-mcp-server [Health: Active]

**Category:** 🔎 Search & Data Extraction  
**Repository:** https://github.com/LinkupPlatform/linkup-mcp-server  
**GitHub Stars:** 29  
**npm Downloads (last month):** 2684  
**Views:** 1  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/linkupplatform-linkup-mcp-server

## Description
Search the web in real time to get trustworthy, source-backed answers. Find the latest news and co…

## Tools
Capabilities this server exposes over MCP:

- **linkup-search** — Search the web in real time using Linkup to retrieve current information, facts, and news from trusted sources. Use this tool for: real-time data (weather, stocks, sports scores, events), breaking news, current events, recent research, product information, up-to-date prices, schedules, and any information not available in your knowledge base. Returns comprehensive content from the most relevant sources.
- **linkup-research** — Submits an autonomous Linkup research task: an agent that investigates the web to answer questions a single search query cannot resolve, returning a synthesized, cited answer. Use for verified answers to precise questions, focused investigations of a defined subject, or broad multi-angle reports. This is async and long-running (can take several minutes): it does not return the final result. It returns immediately with a task "id" and a "status" of "pending". Poll "linkup-get-research" with that id until the status is "completed" (result in "output") or "failed" (reason in "error"). For quick, direct answers prefer "linkup-search" instead.
- **linkup-get-research** — Retrieve the current state of a Linkup research task previously started with "linkup-research". Returns the task object with its "status". While the status is "pending" or "processing", keep polling every few seconds until it reaches a terminal state; polling faster than once per second is rate-limited. When "completed", the result is in the "output" field; when "failed", the reason is in the "error" field.
- **linkup-fetch** — Fetch a URL and return the content of the page. If you are unable to fetch the page content, might be worth trying to render the JavaScript content.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "linkupplatform-linkup-mcp-server": {
    "command": "npx",
    "args": ["-y","linkup-mcp-server"],
    "env": {
      "LINKUP_API_KEY": ""
    }
  }
}
```

**Requires environment variables:** `LINKUP_API_KEY` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What LinkupPlatform-linkup-mcp-server MCP server does

LinkupPlatform-linkup-mcp-server MCP server gives an MCP-compatible AI client access to Linkup's web API. It covers three related tasks: finding current information, conducting longer investigations across multiple sources, and retrieving content from a supplied webpage. Results are intended for questions involving live facts, recent news, prices, schedules, publications, and other information that may have changed since a model's training data.

The server is a fit for agents that need source-backed web results or extracted page content. Search accepts natural-language questions, and the research workflow can return a synthesized answer with citations after investigating multiple sources.

## How it works

LinkupPlatform-linkup-mcp-server can be used as a hosted Streamable HTTP MCP server or run locally over stdio. The hosted endpoint is `https://mcp.linkup.so/mcp`; clients can provide the API key with an `Authorization: Bearer` header, while a query parameter remains available for clients that cannot set custom headers. The local examples invoke the package through `npx` and pass the key as an argument.

A normal search uses `linkup-search`. The `standard` depth is intended for direct questions, while `deep` is intended for analysis across multiple sources. Search requests can limit domains, exclude domains, set publication-date boundaries, cap result count, and optionally include images.

Long-running investigations use two tools. `linkup-research` starts the task and immediately returns an identifier and status; the client then calls `linkup-get-research` until the task is completed or failed. Research can be configured for a precise answer, a focused investigation, or a broader report, with selectable reasoning depth.

## Setup and configuration

Create a Linkup account at the Linkup application site and copy the API key from the dashboard. For a remote setup, configure the hosted MCP URL in the client and supply the key either as a bearer token or through `?apiKey=...`. For a local setup, use the `npx -y linkup-mcp-server` command with the API key argument.

The README provides configurations for Cursor, VS Code, and Codex. It also documents an MCP bundle download for Claude Desktop. The bundle is installed by downloading the `.mcpb` release file, opening it, and entering the API key when prompted.

## Tools and capabilities

LinkupPlatform-linkup-mcp-server exposes these MCP tools:

- `linkup-search` performs real-time web searches using a natural-language query. It supports standard or deep depth, domain inclusion and exclusion, date ranges, result limits, and optional image results.
- `linkup-research` starts an autonomous, multi-source research task. Its options include investigation mode, reasoning depth, domain filters, and date ranges.
- `linkup-get-research` checks the status and retrieves the result of a research task started with `linkup-research`.
- `linkup-fetch` fetches and extracts content from a webpage URL. It can optionally render JavaScript, return extracted images, and include the page's raw HTML.

Use full questions for search and research queries. Choose research when one search request cannot resolve the question or when the answer requires comparison and synthesis across sources.

## Limitations and notes

Research is asynchronous and may take several minutes, so an agent must retain the returned task ID and poll for completion. The API key is required. Remote use depends on an MCP client that supports Streamable HTTP; clients without custom-header support must use the query-parameter form. The available information comes from web pages and search sources, so domain and date filters should be used when a narrower source set is needed.

_Full upstream README: https://allmcps.com/mcp/linkupplatform-linkup-mcp-server/readme_