⚡ Strands Linkup

Linkup web search and fetch tools for Strands Agents. Give your agent real-time, citable web results and clean page content in two lines. 🔗

🌟 Features

• 🔍 linkup_search: real-time web search returning sources (title, URL, content) or a sourced answer.
• 📄 linkup_fetch: fetch any URL as clean markdown, with optional JavaScript rendering.
• 🎚️ Four search depths: from flash (lowest latency) to deep (multi-step agentic search).
• 🧭 Filters: include/exclude domains, date ranges, and result limits.
• ⚡ Async-native: both tools are async @tool functions that run concurrently with the rest of your agent's tool calls.

📦 Installation

pip install strands-linkup

🛠️ Usage

Setting Up Your Environment

1. 🔑 Obtain an API Key: sign up at app.linkup.so to get your API key.

2. ⚙️ Set the API Key: export the LINKUP_API_KEY environment variable before running your agent (or set it with os.environ / a .env file loaded by python-dotenv).

   export LINKUP_API_KEY='YOUR_LINKUP_API_KEY'

Quickstart

from strands import Agent
from strands_linkup import linkup_fetch, linkup_search

agent = Agent(tools=[linkup_search, linkup_fetch])
agent("What are the latest developments in solid-state batteries? Cite your sources.")

The agent uses whichever model you configure for Strands (Amazon Bedrock by default). See the Strands model providers docs to use another provider.

Calling the tools directly

Strands lets you call a tool directly, without going through the model:

result = agent.tool.linkup_search(
    query="Strands Agents SDK release notes",
    depth="standard",
    max_results=5,
)

answer = agent.tool.linkup_search(
    query="Who won the 2026 Tour de France?",
    output_type="sourcedAnswer",
)

page = agent.tool.linkup_fetch(url="https://docs.linkup.so", render_js=False)

📋 Tools

Both tools return the standard Strands tool result shape: {"status": "success" | "error", "content": [{"text": "..."}]}. On success, the text is JSON.

linkup_search

Parameter	Type	Default	Description
query	str	required	The search query, in natural language.
depth	str	"standard"	"flash", "fast", "standard" or "deep".
output_type	str	"searchResults"	"searchResults" for raw sources, "sourcedAnswer" for an answer + sources.
include_domains	list[str]	None	Only return results from these domains.
exclude_domains	list[str]	None	Never return results from these domains.
from_date	str	None	Only results published on or after this date (YYYY-MM-DD).
to_date	str	None	Only results published on or before this date (YYYY-MM-DD).
max_results	int	None	Maximum number of results to return.

Output JSON:

• searchResults: {"results": [{"type": "text", "name": ..., "url": ..., "content": ...}, ...]}
• sourcedAnswer: {"answer": ..., "sources": [{"name": ..., "url": ..., "snippet": ...}, ...]}

Choosing a depth:

• flash: lowest latency, ranked sources and snippets.
• fast: one-shot retrieval in about a second.
• standard: a single pass of agentic search, the right choice for most queries.
• deep: several search iterations for complex, multi-step questions. Slower (can take 5-30s).

linkup_fetch

Parameter	Type	Default	Description
url	str	required	The URL of the web page to fetch.
render_js	bool	False	Render the page's JavaScript before extracting content.

Output JSON: {"url": ..., "markdown": ...}

📚 More Examples

See the examples/ directory for a runnable research agent.

🧑‍💻 Development

make install-dev   # install dependencies and pre-commit hooks
make test          # lint, type-check and run unit tests

🔗 Resources

• Linkup documentation
• Linkup API keys
• Strands Agents documentation
• Strands custom tools

📄 License

MIT. See LICENSE.