Skip to content

Web Search

The Web Search tool searches the web using a SearXNG instance and returns the top results. It gives agents the ability to find current information, research topics, and answer questions that require up-to-date knowledge beyond their training data.

Property Value
Component Type web_search
Category Sub-component (Tool)
Display Name Web Search

Ports

Inputs

Port Data Type Required Description
query STRING Yes Search query string

The query input is provided by the agent's LLM at invocation time.

Outputs

Port Data Type Description
result STRING Formatted list of top search results

Each result includes the title, URL, and a content snippet (up to 200 characters).

Configuration

The Web Search tool accepts the following configuration in extra_config:

Field Type Default Required Description
searxng_url string "http://localhost:8888" Yes Base URL of your SearXNG instance

Example extra_config

{
  "searxng_url": "http://localhost:8888"
}

Prerequisites

The Web Search tool requires a running SearXNG instance. SearXNG is a free, privacy-respecting metasearch engine that aggregates results from multiple search providers.

Quick SearXNG setup with Docker

docker run -d \
  --name searxng \
  -p 8888:8080 \
  -e SEARXNG_SECRET=$(openssl rand -hex 32) \
  searxng/searxng

SearXNG is required

Without a running SearXNG instance, every search call will fail with a connection error. Make sure SearXNG is accessible at the URL configured in extra_config before using this tool.

Usage

  1. Ensure a SearXNG instance is running and accessible
  2. Add a Web Search node from the Node Palette
  3. Connect it to an agent node via the green diamond tools handle
  4. Set the searxng_url in the node's Extra Config field

The agent's LLM can then call the web_search tool with a search query string.

Result format

The tool returns the top 5 results in the following format:

- Result Title
  https://example.com/page
  Content snippet up to 200 characters...

- Another Result
  https://example.com/other
  Another content snippet...

If no results are found, the tool returns No results found.

Example

An agent with a Web Search tool can answer current-events questions:

web_search("Pipelit workflow automation platform")

Returns:

- Pipelit - Visual Workflow Automation
  https://github.com/theuselessai/Pipelit
  Build, connect, and orchestrate LLM-powered agents visually. Design workflows on a React Flow canvas...

- Getting Started with Pipelit
  https://theuselessai.github.io/Pipelit/getting-started/
  Installation guide for the Pipelit visual workflow platform...
flowchart LR
    T[Chat Trigger] --> A[Agent]
    M[AI Model] -.->|model| A
    WS[Web Search] -.->|tool| A

Combine with other tools

Web Search works well alongside HTTP Request. The agent can search for information, then use HTTP Request to fetch full page content or interact with APIs discovered through search results.