> ## Documentation Index
> Fetch the complete documentation index at: https://docs.linkup.so/llms.txt
> Use this file to discover all available pages before exploring further.

# Linkup Skills

> Reusable knowledge modules that teach AI agents how to use Linkup effectively

<video autoPlay muted loop playsInline width="100%" style={{borderRadius: '8px'}}>
  <source src="https://mintcdn.com/linkup-8b5c238e/J-qMLhfHdKUVC8PD/images/linkup-skill-demo.mp4?fit=max&auto=format&n=J-qMLhfHdKUVC8PD&q=85&s=0d1b97515551a147d41282b3638cf97e" type="video/mp4" data-path="images/linkup-skill-demo.mp4" />
</video>

Teach any coding agent to use Linkup well, with one command:

```shell one-click CLI install theme={"system"}
npx skills add LinkupPlatform/skills
```

The installer detects your environment and places the skills where your agent
loads project instructions automatically.

## What are Linkup skills?

Skills are **knowledge for AI agents**, not tools. They are markdown instruction
files that an agent loads when a task matches a skill's purpose, then follows to
make better API calls — choosing the right parameters, writing effective
queries, and avoiding common mistakes. They work in any agent environment that
supports instruction files. Learn more about the [Skill CLI](https://skills.sh/).

The Linkup skills are generated from
[**linkup-for-agents**](https://github.com/LinkupPlatform/linkup-for-agents),
the full Linkup knowledge pack with knowledge files and 18 ready-to-adapt
workflow recipes.

## What's included

The package installs five self-contained skills:

| Skill             | Use it for                                                                                                 |
| ----------------- | ---------------------------------------------------------------------------------------------------------- |
| `linkup-search`   | Any web lookup or research query — the default. Covers `depth`, output types, and query-as-retrieval-plan. |
| `linkup-fetch`    | Reading one known URL as clean Markdown.                                                                   |
| `linkup-research` | Minutes-long, multi-source investigations via `/v1/research`.                                              |
| `linkup-extract`  | Bulk structured rows from one listing page via `/v1/extract`.                                              |
| `linkup-workflow` | Turning a business goal into a multi-step Linkup workflow, with 18 bundled recipes.                        |

Together, the skills replace generic tool descriptions with structured guidance:
how to pick `depth`, choose an output type, phrase queries as retrieval plans,
decide between scraping inside
[**Search**](/pages/documentation/endpoints/search/overview) and the
[**Fetch**](/pages/documentation/endpoints/fetch/overview) endpoint, and handle
advanced cases like LinkedIn extraction and search-then-scrape chains. Each
skill also bundles the reference material it needs in a local `references/`
directory.

For the full detail, follow the links: query writing lives in
[Search best practices](/pages/documentation/endpoints/search/best-practices),
workflow patterns and recipes in
[For AI agents](/pages/documentation/get-started/for-agents), and client setup
in the [MCP guide](/pages/integrations/mcp/mcp).

## Prerequisites

Before installing the skills, make sure you have:

* A [Linkup API key](https://app.linkup.so).
* Linkup connected to your agent through the
  [MCP server](/pages/integrations/mcp/mcp), an
  [SDK](/pages/sdk/python/python), or the
  [raw API](/pages/documentation/get-started/quickstart).

## Where it works

The skills install into your project and load automatically when an agent handles
web search, content extraction, research, or workflow planning via Linkup — no
manual invocation needed.

| Environment                | How it loads                         |
| -------------------------- | ------------------------------------ |
| **Claude Code**            | Automatically when relevant          |
| **Cursor**                 | Via `.cursor/rules/`                 |
| **Windsurf**               | Via `.windsurfrules`                 |
| **Cline / GitHub Copilot** | From project-level instruction files |

## How it works

When an agent hits a task involving web search, company research, or content
extraction:

1. It detects access to Linkup tools (via MCP or SDK).
2. The matching Linkup skill instructions load into context.
3. The agent follows the skill's reasoning to choose `depth`, query style, and `outputType`.
4. The result: more accurate queries, fewer wasted API calls, better results.

## Example

Without the skill, an agent might write:

```
query: "Tell me about Datadog"
depth: "standard"
```

With `linkup-search`, it reasons through the framework and writes:

```
query: "Find Datadog's pricing page. Scrape it. Extract plan names, per-host prices, and included features for each tier."
depth: "deep"
```

The skills teach agents to be specific about what to retrieve, where to look,
and how to chain steps — the way an experienced developer uses the API.

## Resources

* [GitHub repository](https://github.com/LinkupPlatform/skills): installable Linkup skills.
* [linkup-for-agents](https://github.com/LinkupPlatform/linkup-for-agents): source knowledge pack and workflow recipes.
* [For AI agents](/pages/documentation/get-started/for-agents): the agent hub — drop-in prompt, workflow patterns, and endpoint briefs.
* [Search best practices](/pages/documentation/endpoints/search/best-practices): detailed query-writing guidance.
* [MCP setup](/pages/integrations/mcp/mcp): connect Linkup to your agent via MCP.

<Info>
  **Need help?** Email `support@linkup.so`, ping us on [Discord](https://discord.com/invite/9q9mCYJa86), or [talk to us](https://calendar.app.google/zNUZz7RxgMMk9pKW7).
</Info>