Use case
Web research agent with a search tool
Some questions need live data. Connect Tavily once and attach its web search to an agent as a tool. The agent decides when to search, with which query, and how many times.
How the workflow runs
- The trigger carries a research question.
- The research agent has one tool: a Connector node running Tavily · Web search, connected to its
toolsport with a dashed tool edge. - The agent calls the tool as often as it needs, up to its max tool rounds (6 by default), then writes the summary.
- The Output node marks the summary as the workflow result.
Describe the tool
A connector action already describes its inputs to the model. Leave Query empty so the agent fills it, and fix the rest so every call stays within bounds:
Query: (left empty, the agent fills it)
Depth: basic
Results: 5
Tool description:
Search the web. Call once per distinct question.Searching your own index instead? Add it under APIs with a saved GET /search?q={query} endpoint and attach an API node the same way.
When a call fails
A failing tool does not fail the run. The agent receives an error result and can retry with different arguments or explain what it could not find. Every call and response is recorded on the run for replay.
Keep it bounded
Keep reading
- Connector node: run an action on a connected appThe Connector node runs a GitHub, Slack, Notion, Linear or Tavily action with your account's connection, as a workflow step or as a tool an agent can call.
- Connectors: GitHub, Slack, Notion, Linear and TavilyConnect GitHub, Slack, Notion, Linear and Tavily to your Worfilo account, then run their actions as workflow steps or give them to an agent as tools.
- Execution controls: retries, timeouts and error handlingControl how Worfilo runs AI workflows: per-node retries with backoff, timeouts and error modes, and run-wide parallelism, timeouts and token caps.
Build it on the canvas
Create a free account, describe the workflow or wire it yourself, and run it in the browser.