Search Tools

The Search provider performs a web search and returns the most relevant results, each with a title, link, snippet, and date. Use it in a workflow when an agent or downstream block needs fresh, real-world information from the web rather than relying on model knowledge alone.

Overview

Operations

Configuration

search_tool

This tool takes no API key — it posts to the platform-internal /api/tools/search endpoint, which handles the upstream search credentials and billing. There are no secret parameters to configure.

Outputs

search_tool

• results (array) — The list of search results. Each entry contains:
  • title (string) — The result title.
  • link (string) — The URL of the result.
  • snippet (string) — A short text excerpt from the result.
  • date (string, optional) — The publication or indexed date of the result, when available.
  • position (number) — The rank/position of the result in the result set.
• query (string) — The search query that was executed.
• totalResults (number) — The total number of results found.

YAML Example

search_1:
  type: search
  name: "Search"
  inputs:
    operation: "search_tool"
    query: "latest advances in AI agents 2026"
  connections:
    outgoing:
      - target: next-block-id

Tips

• No credentials are required in your workflow. The block calls the internal /api/tools/search route, so you do not provide an API key — the platform manages the upstream search provider and billing.
• Each search costs $0.01 per query (per the block definition), so avoid running it in tight loops; cache or reuse results where possible.
• Pass dynamic values into query with double-brace references, e.g. query: "{{agent_1.content}}", to search based on a previous block's output.
• Downstream, iterate over {{search_1.results}} (an array of title/link/snippet/date/position) rather than assuming a fixed result count — use {{search_1.totalResults}} to know how many were returned.